Novel nano materials for high performance logic and memory devices

Saptarshi Das, Purdue University

Abstract

After decades of relentless progress, the silicon CMOS industry is approaching a stall in device performance for both logic and memory devices due to fundamental scaling limitations. In order to reinforce the accelerating pace, novel materials with unique properties are being proposed on an urgent basis. This list includes one dimensional nanotubes, quasi one dimensional nanowires, two dimensional atomistically thin layered materials like graphene, hexagonal boron nitride and the more recently the rich family of transition metal di-chalcogenides comprising of MoS2, WSe2, WS2 and many more for logic applications and organic and inorganic ferroelectrics, phase change materials and magnetic materials for memory applications. Only time will tell who will win, but exploring these novel materials allow us to revisit the fundamentals and strengthen our understanding which will ultimately be beneficial for high performance device design. While there has been growing interest in two-dimensional (2D) crystals other than graphene, evaluating their potential usefulness for electronic applications is still in its infancies due to the lack of a complete picture of their performance potential. The fact that the 2-D layered semiconducting di-chalcogenides need to be connected to the "outside" world in order to capitalize on their ultimate potential immediately emphasizes the importance of a thorough understanding of the contacts. This thesis demonstrate that through a proper understanding and design of source/drain contacts and the right choice of number of MoS2 layers the excellent intrinsic properties of this 2D material can be harvested. A comprehensive experimental study on the dependence of carrier mobility on the layer thickness of back gated multilayer MoS 2 field effect transistors is also provided. A resistor network model that comprises of Thomas-Fermi charge screening and interlayer coupling is used to explain the non-monotonic trend in the extracted field effect mobility with the layer thickness. The non-monotonic trend suggests that in order to harvest the maximum potential of MoS2 for high performance device applications, a layer thickness in the range of 6-12 nm would be ideal. Finally using scandium contacts on 10nm thick exfoliated MoS2 flakes that are covered by a 15nm ALD grown Al2O3 film, record high mobility of 700cm2/Vs is achieved at room-temperature which is extremely encouraging for the design of high performance logic devices. The destructive nature of the readout process in Ferroelectric Random Access Memories (FeRAMs) is one of the major limiting factors for their wide scale commercialization. Utilizing Ferroelectric Field-Effect Transistor RAM (FeTRAM) instead solves the destructive read out problem, but at the expense of introducing crystalline ferroelectrics that are hard to integrate into CMOS. In order to address these challenges a novel, fully functional, CMOS compatible, One-Transistor-One-Transistor (1T1T) memory cell architecture using an organic ferroelectric – PVDF-TrFE – as the memory storage unit (gate oxide) and a silicon nanowire as the memory read out unit (channel material) is proposed and experimentally demonstrated. While evaluating the scaling potential of the above mentioned organic FeTRAM, it is found that the switching time and switching voltage of this organic copolymer PVDF-TrFE exhibits an unexpected scaling behavior as a function of the lateral device dimensions. The phenomenological theory, that explains this abnormal scaling trend, involves in-plane interchain and intrachain interaction of the copolymer - resulting in a power-law dependence of the switching field on the device area (ESW α ACH0.1) that is ultimately responsible for the decrease in the switching time and switching voltage. These findings are encouraging since they indicate that scaling the switching voltage and switching time without aggressively scaling the copolymer thickness occurs naturally while scaling the device area – in this way ultimately improving the packing density and leading towards high performance memory devices.

Degree

Ph.D.

Advisors

Appenzeller, Purdue University.

Subject Area

Electrical engineering|Nanoscience|Materials science

COinS