|Thesis title:||Modeling of Conduction and Reliability Characteristics for Phase Change Non-Volatile Memories|
|Research area:||Microelectronics and Emerging Technologies|
Exploiting the ever increasing market demand for portable consumer products requiring permanent and high-density data storage, semiconductor non-volatile memories (NVMs) have gained in the last decades an explosive success. Music players, digital cameras, USB drives, cell phones and the emerging Solid State Disks (SSDs) are only a few examples of the ubiquitousness and the ever increasing part got by non-volatile memory in reshaping our lifestyle. In this play the leading role has been performed by Flash memory, that from a simple concept in the early 80’s grew up and generated close to $23 billion in worldwide revenue in 2007, undoubtedly representing the actual mainstream memory on the market.
This enormous success was essentially driven by Moore’s Law, that lead to dramatic reductions in unit cost and to the creation of new fruitful markets. The ever increasing demand for more memory bits largely repaid the continuous efforts devoted for manufacturing memory chips with increased performance and functionality, resulting in a sort of virtuous circle. Thus, despite their higher cost per bit with respect to magnetic hard disk drives, semiconductor memories resulted the winning solution in all the consumer products requiring light weight, low size, low power consumption and high reliability.
However, a further increase in storage capacities with a simultaneous cost per bit reduction is now mined by the physical and technological constraints of conventional memory technology and will thus require for next technology nodes something more than the scaling of feature size. As a result, to provide a better trade-off between scalability and reliability as well as the basis for the evolution of the actual storage hierarchy in its whole, new memory concepts have been recently object of intense investigation. More interestingly, most of these solutions have renewed the attention to certain classes of materials whose nature is far from the regular perfection that had permeated as a condicio sine qua non the outstanding evolution of microelectronics so far. Chalcogenides alloys, binary/ternary oxides, as well as other amorphous semiconductor materials exploited in these new kind of memory devices, can in fact be cited as representative examples of the materials and design revolution felt by the semiconductor industry in its whole. In this frame, among the proposed emerging concepts, chalcogenide-based phase-change memory holds a privileged position due to the good degree of technology maturity, supported by its promising scaling potential and a broad application range.
PCM working principle is essentially different from conventional floating gate devices, where the information is associated to an amount of stored charge: It relies on a change of the resistance in the active material, hence its classification as a resistive memory. Such a functionality is obtained by smartly exploiting the peculiarities of particular materials, usually chalcogenide alloys (e.g. Ge2Sb2Te5), that can be reversibly switched between crystalline and amorphous phase by electrically induced phase transitions.
Given the resistive type of PCM, the transport phenomenology contributes in a fundamental way to determine the technology performance. The exploration and modeling of conduction plays therefore a central role, allowing to accurately predict the read-window budget between set and reset states and to evaluate device functionality as a function of different programming and operating conditions – such as temperature, time, voltage and current –, providing support for the technology development.
The Ph.D Dissertation will address the conduction properties and the reliability characteristics of PCM devices by both experimental and numerical investigations. New interesting phenomena for active volumes of decananometric sizes will be reported, including: (i) an anomalous dependence of the resistance on the thickness of the programmed amorphous region, (ii) current localization effects related to the non-Arrhenius behavior of the resistance in temperature, (iii) the occurrence of Lorentzian components (Random Telegraph-signal Noise - RTN) in the current spectra and (iv) experimental evidence of how drift and noise phenomena are both driven by temporal variations of the energy for hopping within the material.
In order to address these phenomena, a new physics-based framework for the conduction in amorphous phase, namely the distributed Poole-Frenkel (DPF) model, will be introduced. Transport will be described in terms of Poole-Frenkel (PF) conduction through localized states with distributed energy barriers for hopping, leading to a more general form of conduction with respect to previously reported models for PCM devices. Also relying on previous works on amorphous solids and percolation in disordered materials, the conduction process will be reformulated and numerically implemented in terms of an equivalent Random Resistance Network (RRN) with an exponentially wide spectrum of local resistances. Through the inclusion of the physical insights coming from the observed experimental dependences, the proposed approach will be shown to allow for a quantitative description of the temperature and time dependence of the programmed state, therefore contributing to a significant progress in the understanding of its properties. Finally, the model will be adopted to address some reliability scaling projections in terms of the resistance window for future technology nodes. These reliability investigations represent the key for the qualification of phase-change memory for both single bit and multilevel applications.