Evaluation of Different High Κ Materials for In situ Growth of Carbon Nanotubes

Abstract

Carbon nanotubes (CNTs) exhibit intrinsic electrical properties superior to those of silicon. Working field-effect devices with CNTs as channel material (CNTFETs) have first been demonstrated on research level by Tans and coworkers [1]. Although progress has been achieved in fabricating devices on a larger production scaled level [2], a fully satisfying process is still missing. By now, processes are either tedious or yield low performing CNTs. There are two main possibilities to incorporate CNTs in electronic devices: either make use of ex situ grown CNTs or grow them in situ. With the ex situ approach CNTs are grown by a separate process, transferred into a solution, purified from contaminants and from CNTs with undesired properties, and then finally deposited onto the target substrate for further processing. The advantage of this approach is that it yields a population of CNTs which have selectable properties and a scalable density. When using the in situ approach CNTs are grown directly in their final place, thus there is no need for any transferring processes. On the other hand, in situgrowth tends to yield a CNT population with all types of chirality and therefore very diverse electrical properties (semi-conducting with wide or narrow bandgap or even metallic behavior). As shown in [3] field-effect devices on the basis of in situ grown CNTs can be fabricated if a process yields a high density CNT population with small diameters; this makes selective destruction of non-performing, i. e. metallic CNTs by electrical burn pulses possible [4]. In that manner, the disadvantage of in situgrowth to also produce CNTs with undesired properties is overcome. The technological key to this is a process and material system that produces a high amount of thin CNTs [5]. Based on a process which is described in detail in [6] we investigate the growth of CNTs on different high κstacks. CNT growth is done by means of catalytic chemical vapor deposition (CCVD) from methane and a catalyst. As catalyst < 1 nm of e-beam evaporated nickel is used with an optional underlayer of 5 nm aluminum [3, 5]. As substrates we use silicon wafers with 30 nm of thermally grown stress oxide for all samples. The following material stacks are prepared (with the deposition method given in brackets): silicon dioxide (dry oxidation) / aluminum/nickelsilicon nitride (chemical vapor deposition) / silicon dioxide (wet oxidation) / aluminum/nickelaluminum oxide (atomic layer deposition; ALD) / nickelaluminum oxide (reactive magnetron sputtering [7]) / nickelaluminum nitride (reactive magnetron sputtering [8]) / nickel The results are monitored using atomic force microscopy (AFM) and are shown in fig. 1 to 5. While CNTs are grown on all samples, stack 3 shows the most promising result in terms of CNT amount and diameter. Electrical characterization (fig. 6) of as-grown CNTs on stack 3 show PMOS behavior with an current drive of 140 µA and an on/off ratio of only a factor of 2.9. By applying an electrical burn pulse to this device the on/off ratio is improved to 4 orders while the current drive is only decrease by 2 orders of magnitude to 5 µA. (We will report on the used "burn pulse" apparatus elsewhere.) We aim to improve the current drive by using a UV assisted ozone treatment before evaporation of the contact metal to realize a better CNT-metal contact. Acknowledgments The authors would like to thank Dr. Barbara Abendroth for the possibility to prepare the aluminum oxide films by ALD in the facilities of Technische Universität Bergakademie Freiberg.