Characterization of Trapping Charges and Contact Properties in Carbon Nanotube Transistors

Abstract

Carbon nanotubes (CNTs) have received tremendous attention due to unique electrical properties because of their one-dimensional (1D) material nature. Owing to extraordinary thermal conductivity, mechanical, and electrical properties, CNTs have been considered as the candidate material for next-generation electronics. Among various applications, CNTs as semiconducting layers in transistors are the most promising, since silicon technology is expected to reach its performance limits soon, and the demand for flexible/transparent electronics is high. Although the unique 1D feature of CNTs is beneficial in terms of flexibility and tuning electrical properties for the required device operation, this 1D feature also yields some issues. (i) transistor operation could largely be influenced by charge diffusion from the CNTs to the surrounding dielectric. This causes the electrical performance of CNT transistors to be largely influenced by the trapping of charges. (ii) The effective contact area is smaller than the geometrically defined channel to the source/drain (S/D) contact region, since the diameter of a CNT is typically only 1–3 nm. Thus, in this thesis, two important interfaces, the CNT/gate insulator and the CNT/S/D contact, are discussed. First, the correlation between charge trapping and the charge transport properties in random-network single-walled carbon nanotube (SWCNT) transistors was investigated using direct current (DC) and transient analysis. DC analysis was conducted throughout the temperature-dependent forward (12 V to -12 V) and reverse (- 12 V to 12 V) gate sweep. The activated energy (Ea) extracted from the temperature-dependent mobility showed that the charge transport in our SWCNTs is not governed by the residual ions or the defects in the gate dielectric layer. Further investigation was conducted by extracting the temperature-dependent charge carrier density (n) and trap density (Nt). The charge carrier density (n) and trap density (Nt) showed similar temperature-dependent behavior, which indicates that the charge trapping and hysteric behavior in SWCNT transistors is primarily from the charge injection from the CNTs to the surrounding dielectric. Subsequently, transient measurement was carried out for further investigation. The transient measurement was performed with a small load resistor (Rload), such that the measurement circuit was small enough to extract the intrinsic RC time constant (τ) value of the SWCNTs channel. To investigate the effect of charge trapping on the transient response, an empirical equation was developed based on the theoretical trapping model. The transient mobility (μ_tr) showed similar temperature-dependent trends with the mobility (μ) extracted from the DC analysis, which further supports that the main factor of charge trapping in SWCNTs is not the residual ions, or the defects in the gate dielectric layer. Throughout the empirical equation, the charge velocity distribution in SWCNTs was successfully explained by the trapping of charges. The correlation between charge carrier density (n) and intrinsic time constant difference (Δτ) was also investigated. Throughout this analysis, we confirmed that the transient response is significantly influenced by charge trapping. We also found that the charge transport in the SWCNTs channel is largely influenced by the shallow traps rather than the deep traps. The trapping and detrapping rates were also extracted from the transient analysis. Second, the effect of the graphene S/D contact on the electrical performance of the SWCNT transistor was investigated. The contact resistance between SWCNTs and the S/D electrode can be improved by forming a large number of junctions between the SWCNTs and the S/D electrode. Thus, forming a dense SWCNT film is an effective way to increase the junctions between the SWCNT film and the S/D electrode. However, too dense an SWCNT film can cause the CNTs to form bundles and results in the decrease in the on/off current ratio (Ion/Ioff). Thus, a trade-off relationship exists between the contact resistance and the Ion/Ioff. To overcome this trade-off relationship, we employed graphene as the S/D electrode for the SWCNT transistor. The bottom-gate bottom-contact (BGBC) geometry was selected for the graphene S/D contact SWCNTs (Gr-SWCNTs) transistor such that the drain current (ID) could additionally be modulated by the graphene layer. A palladium (Pd) S/D contact SWCNTs (Pd-SWCNTs) transistor with the same device geometry was also fabricated for comparison. To determine whether graphene formed by chemical vapor deposition (CVD) is suitable for the S/D electrode, our graphene film was characterized thoroughly by the optical microscopic image, atomic forced microscopic (AFM) image, scanning electron microscope (SEM), and the Raman spectra. Throughout the graphene film analysis, we found that our monolayer graphene sheet consists of the combination of large-sized grains (diameter of grain ~100 μm) and small-sized grains (~1 μm). The transmission line method (TLM) results show that the resistivity of graphene was small enough to be used as the S/D electrode in a single transistor device. The effective work function investigation result indicates that the work function of graphene is well aligned to the work function of Pd. Since the surface energy of the underlying films could influence the deposition of the SWCNTs film, contact angle measurements were performed to acquire the surface energy of each layer. The resulting surface energy was then systematically compared with the AFM image of SWNCTs in the channel. The results showed that the graphene S/D contact is favorable for the SWCNTs to be densely formed in the channel, presumably due to the selective wetting properties that led the SWCNTs to be well confined in the channel. The transfer characteristics of the Gr-SWCNT transistor showed that high mobility (μ) with good Ion/Ioff could be achieved by employing the graphene S/D contact. The conducting behavior and the influence of the contact resistance to the electrical performance of the transistor were further investigated using the stick percolating system and the TLM method. Both Gr-SWCNT and Pd-SWCNT transistors showed the exponent of the stick percolating system near 1 (m = 1). The TLM results showed that the contact resistance of Gr-SWCNTs was lower than that of Pd-SWCNTs.