MRS Meetings and Events


EL21.12.02 2023 MRS Spring Meeting

Assessing the Impact of Ti Doping and MAX-Phase Ti2AlN Electrode in HfOx Neuromorphic Devices

When and Where

Apr 13, 2023
3:45pm - 4:00pm

Moscone West, Level 3, Room 3011



Fabia Farlin Athena1,Moses Nnaji1,Matthew P. West1,Samuel Graham2,Eric Vogel1

Georgia Institute of Technology1,University of Maryland2


Fabia Farlin Athena1,Moses Nnaji1,Matthew P. West1,Samuel Graham2,Eric Vogel1

Georgia Institute of Technology1,University of Maryland2
Oxide-based non-volatile memories have great potential for neuromorphic computing applications.<sup>1</sup> In HfO<sub>x</sub>filamentary memory, when bias is applied, a conductive filament (CF) is formed, within which O ion motion in and out causes rupture and reformation of the CF, and this process primarily governs the switching. However, challenges such as high forming voltage (V<sub>F</sub>), abrupt resistance change, and high off-state current prohibit its widespread adoption.<sup>2</sup><br/><br/>In this study, we independently assess the impact of Ti doping and electrode material optimization on HfO<sub>x</sub> devices with the goal of overcoming the problems in HfO<sub>x</sub> filamentary memory. The Ti-doped HfO<sub>x</sub> devices were fabricated, where the ~5 nm oxide was deposited via atomic layer deposition. Electrical characterization shows that the <b>V<sub>F </sub>decreases</b> (~2.1 V) with the increase of Ti doping.<sup>3</sup> In addition, analog pulse measurements show that a 25% Ti doping <b>improves the linearity in resistance change</b> during the reset. However, the switching is not substantially impacted, and the <b>off-current</b><b> is high</b>.<sup>4</sup> A few O ion motion at the CF break region governs the switching. We determined the location of the break via a direct tunneling electron transport model and a finite element analysis. Both analyses show that the break is close to the reset anode electrode; thus, the electrode has a significant impact on the switching. We hypothesize that a reduction in thermal conductivity (K) of the electrode will lower the off-current. A low electrode will cause less heat removal and more oxygen ion motion, thus will result in a thicker break in CF and less off-current. However, the challenge is that most of the electrodes have high K (e.g., K<sub>Au</sub> = 310 W/m.K., K<sub>TiN</sub> = 29 W/m.K). Ti<sub>2</sub>AlN, like many other MAX phase materials, has unique metallic (high electrical conductivity) and ceramic properties (low thermal conductivity, K<sub>MAX phase </sub>~4.63 W/m.K) as it contains alternating layers of metallic and covalent bonds.<sup>5 </sup> Thus, HfO<sub>x </sub>devices with Ti<sub>2</sub>AlN MAX phase as the bottom electrode were fabricated. Alternating layers of Ti and AlN were first deposited at room temperature via magnetron sputtering and then thermally annealed to obtain Ti<sub>2</sub>AlN. Raman spectroscopy of the electrode showed the characteristic spectra of Ti<sub>2</sub>AlN. The HfO<sub>x</sub> devices with MAX phase BE show a <b>one-order-of-magnitude improvement in the off-current</b> (HRS ~ 450 kΩ), a large switching window (~200), good endurance (&gt;100) and retention (~10<sup>4 </sup>s) of states. Furthermore, analog pulsing measurements are conducted to analyze the synaptic behavior of the devices. A COMSOL Multiphysics<sup>®</sup> model is used to develop a fundamental understanding of the observed results.<br/>Overall, this study demonstrates that through optimization of the materials in HfO<sub>x</sub> memory, the electrical responses can be tailored for neuromorphic application.<br/><br/><br/>1. Yang, J. J., et al. <i>Nature nanotechnology</i>, <i>8</i>(1), 13-24.<br/>2. K. Moon, et al., <i>Faraday discussions</i> <b>213</b>, 421-451 (2019).<br/>3. Athena, F. F., et al. <i>Journal of Materials Chemistry C</i>, <i>10</i>(15), 5896-5904.<br/>4. Athena, F. F., et al. <i>Journal of Applied Physics</i>, <i>131</i>(20), 204901.<br/>5. Zhang, T., et al. <i>J. Ceram. Process. Res</i>, <i>13</i>, S149-S153.<br/><br/><br/><b>Acknowledgements: </b>This work was supported by the AFOSR MURI under Award No. FA9550-18-1-0024 to Eric M. Vogel. In part at the Georgia Tech Institute for Electronics and Nanotechnology, a member of the NNCI, which is supported by the NSF (ECCS-2025462). This material is based upon the work supported by the 2022-23 IBM PhD Fellowship Award to Fabia F. Athena.


atomic layer deposition

Symposium Organizers

Iuliana Radu, Taiwan Semiconductor Manufacturing Company Limited
Heike Riel, IBM Research GmbH
Subhash Shinde, University of Notre Dame
Hui Jae Yoo, Intel Corporation

Symposium Support

Center for Sustainable Energy (ND Energy) and Office of Research

Raith America, Inc.

Publishing Alliance

MRS publishes with Springer Nature