IntroductionHfO 2 gate dielectrics are considered to be the most promising high-k dielectrics to meet the future ULSI application, due to its high dielectric constant and excellent thermal stability [1][2][3]. Besides, as MOS devices continue to be scaled down, the low supply voltage is desirable to minimize the power consumption. Dynamic threshold voltage (DT) MOSFET has been extensively studied [4], being attractive for lower power supply voltage applications. In addition, the CESL stressor [5][6] has been used to improve mobility in CMOS technology. However, the combination of DT and CESL nMOS with high-k in CMOS technology has yet to be proposed. In this work, the mechanism and reliability of DT-CESL-HfO 2 nMOSFETs were deeply investigated.
Device PreparationA 0.35-μm process was used with local-oxidation-of-silicon (LOCOS) isolation. After standard RCA cleaning, the HfO 2 thin film with 1.6 nm (CET) was deposited on Si wafer. Then, α-Si (deposition at 550 o C,) of thickness 50 nm was deposited in order to get the local tensile strain n-channel [6], and the final poly-gate thickness was 150 nm. After the S/D formation, a low pressure chemical vapor deposition (LPCVD) 50 nm silicon nitride (CESL) was directly deposited on the transistor at 780 o C and followed by a 200 nm PE-SiO 2 deposition. After contact alignment, the PE-oxide and SiN layer on the S/D regions were etched in the same system. After these processes, a four-level metallization (Ti-TiN-Al-TiN) was carried out in the PVD system for the contact. Figure 1 shows the schematic diagram of DT-CESL-nMOS transistor with HfO 2 gate dielectric. As shown in Fig. 1, the poly-Si gate is directly connected to the p-well region. Figure 2 shows the I D -V D characteristics of the CESL strained HfO 2 nMOSFETs at normal and DT biasing, where the device channel length and width were 0.5 and 10 μm, respectively. The drain current of DT mode showed 133% increase over the normal mode for the CESL-device at V D = V G = V B = 0.7 V, as shown in Fig. 2. This large increase can be attributed to the body bias effect. For the DT mode, the device is in on-state. On the contrary, the device with normal operation is operated in depletion region. Figure 3 demonstrated the transconductance (g m ) of CESL-devices with different channel length, at DT and normal biasing, respectively. The g m under DT mode is larger than that under the normal mode. The enhancement of g m , compared to the normal operation mode, was increased, with channel length decreasing as shown in this figure. A roughly 138 % increase can be observed for the 0.35μm CESL-device at DT biasing, by comparison with the normal mode. On the other hand, the increase of g m was only about 40 % for the 1μm CESL-device. These results imply that the increase of electron mobility is different in different channel length for the DT-CESL-MOS with HfO 2 gate dielectrics. In addition, this performance enhancement of the DT-CESL-nMOS can be speculated to the inversion charge density increase at DT biasing, as shown in Fig. 4. T...