David Cabrera


As semiconductor devices continue to scale downward, and portable consumer electronics become more prevalent there is a need to develop memory technology that will scale with devices and use less energy, while maintaining performance. One of the leading prototypical memories that is being investigated is phase change memory. Phase change memory (PCM) is a non-volatile memory composed of 1 transistor and 1 resistor. The resistive structure includes a memory material alloy which can change between amorphous and crystalline states repeatedly using current/voltage pulses of different lengths and magnitudes. The most widely studied PCM materials are chalcogenides - Germanium-Antimony-Tellerium (GST) with Ge2Sb2Te3 and Germanium-Tellerium (GeTe) being some of the most popular stochiometries. As these cells are scaled downward, the current/voltage needed to switch these materials becomes comparable to the voltage needed to sense the cell's state. The International Roadmap for Semiconductors aims to raise the threshold field of these devices from 66.6 V/μm to be at least 375 V/μm for the year 2024. These cells are also prone to resistance drift between states, leading to bit corruption and memory loss.

Phase change material properties are known to influence PCM device performance such as crystallization temperature having an effect on data retention and litetime, while resistivity values in the amorphous and crystalline phases have an effect on the current/voltage needed to write/erase the cell. Addition of dopants is also known to modify the phase change material parameters.

The materials G2S2T5, GeTe, with dopants - nitrogen, silicon, titanium, and aluminum oxide and undoped Gallium-Antimonide (GaSb) are studied for these desired characteristics. Thin films of these compositions are deposited via physical vapor deposition at IBM Watson Research Center. Crystallization temperatures are investigated using time resolved x-ray diffraction at Brookhaven National Laboratory. Subsequently, these are incorporated into PCM cells with structure designed as shown in Fig.1. A photolithographic lift-off process is developed to realize these devices. Electrical parameters such as the voltage needed to switch the device between memory states, the difference in resistance between these memory states, and the amount of time to switch are studied using HP4145 equipped with a pulsed generator. The results show that incorporating aluminum oxide dopant into G2S2T5 raises its threshold field from 60 V/μm to 96 V/μm, while for GeTe, nitrogen doping raises its threshold field from 143 V/μm to 248 V/μm. It is found that GaSb at comparable volume devices has a threshold field of 130 V/μm. It was also observed that nitrogen and silicon doping made G2S2T5 more resistant to drift, raising time to drift from 2 to 16.6 minutes while titanium and aluminum oxide doping made GeTe drift time rise from 3 to 20 minutes. It was also found that shrinking the cell area in GaSb from 1 μm2 to 0.5 μm2 lengthened drift time from 45s to over 24 hours.

The PCM process developed in this study is extended to GeTe/Sb2Te3 multilayers called the superlattice (SL) structure that opens opportunities for future work. Recent studies have shown that the superlattice structure exhibits low switching energies, therefore has potential for low power operation.

Library of Congress Subject Headings

Nonvolatile random-access memory--Materials; Chalcogenides--Electric properties

Publication Date


Document Type


Student Type


Degree Name

Electrical Engineering (MS)

Department, Program, or Center

Electrical Engineering (KGCOE)


Santosh K. Kurinec

Advisor/Committee Member

Robert E. Pearson

Advisor/Committee Member

Ivan Puchades


Physical copy available at RIT's Wallace Library at TK7895.M4 C33 2014


RIT – Main Campus

Plan Codes