No abstract

layer thickness is to be optimized by balancing the Vt Ultra thin layers of Magnesium containing cap modulation and accompanying mobility loss. All these stacks layers are deposited on Hf based dielectrics prior to exhibit at least 4 orders of magnitude lower leakage deposition of the TiN/Poly-Si electrode stack [1] to achieve compared to SiON/Poly stacks (Fig. 5). Inversion Split-CV, Band-edge (BE) high-wiMetal nMOSFETs with good Id-Vg, and mobility curves for nFET stacks with HfSiO mobility (190 cm2/Vs @ IMV/cm) at Ti,1 (1.45 nm), in a gate dielectric and different cap thicknesses are shown resectivley first process flow. It is shown that Vt can be modulated in Figs 6-8. Compared to HfO2, Tinv and Vt modulation anywhere between midgap and band edge by changing the behavior are the same for HfSiO stacks but there is a greater cap layer thickness. Short channel devices with band edge mobility degradation. Mobility degrades by almost 25% for characteristics are demonstrated down to 40 nm. HfSiO based BE nFET stacks as compared to 10% in HfO2 Introduction stacks. Although engineering solutions have been shown for the Cross-sectional HRTEM of BE stacks with cap M2 (Fig. implementation of High-K/Poly-Si stacks in recent times for 9) does not show a distinct layer between the HfO2/TiN low power applications [2], high-k/metal electrode gate interface suggesting that the cap layer has completely stacks with BE workfunctions still remain a necessity to meet intermixed with the HfO2 layer. EELS/EDX analysis reveals the high performance CMOS technology requirements for the presence of Mg in the HfO2 layer. Backside SIMS data 45nm and beyond nodes. Finding suitable metal gates withfrom these stacks confirms that there is no Mg in the appropriate nFET and pFET workfunctions that are stable in underlying silicon channel (Fig. 10). Based on materials a gate first process flow (highest TAIOOOC) has remained a analysis the mechanism causing the Vt shifts with Mg challenge, though high electron mobility [3,4] and low charge containing cap layers is believed to be similar to that seen trapping [5] have been demonstrated by using a mid-gap TiN with La containing cap layers [9]. The Vt shifts can either be electrode in conjunction with Hf-based dielectrics. Oxygen attributed to built-in dipole layers in the dielectric layer when vacancy induced positive shifts have been postulated to be the cap layer intermixes with HfO2 or the generation of the reason for high Vt in HfO2/p-metal stacks [6]. But for oxygen vacancies by aleovalent substitution of Mg for Hf nFET -the challenge has been in finding a thermally stable resulting in a net positive charge (MgO = Mg" + o + V) metal electrode with the right workfunction that is compatible c M ± with a gate first self-aligned process. Good progress has been Finally, Figure 11 shows Vt roll-off curves down to made in the last 2 years towards this and the more attractive channel lengths of 40nm, demonstrating that the Vt shifts results include a gate last solution [7] and a gate first solution ...

A 32 nm BEOL with PVD CuMn seedlayer and conventional PVD-TaN/Ta liner was fully characterized by fundamental, integrated, and reliability methods. CuMn was confirmed to have fundamental advantages over CuAl, such as higher electromigration (EM) reliability for the same Cu line resistance (R). Both low R and high reliability (EM, SM, and TDDB) were achieved. Improved extendibility of CuMn relative to CuAl was also supported by studies of alloy interactions with advanced liner materials Ru and Co, and by enhancement of ultra-thin TaN barrier performance.

Many commercial relational data base systems provide two join methods: 1) Nested Loop join and 2) Sort Merge join. Nested Loop join exploits indexes on the inner table's join column. Sort Merge join benefits from the efficiency of bulk sequential disk accesses. Both provide good performance when selected correctly by the query optimizer, but could prove prohibitively expensive if selected incorrectly. We describe a new method called Hybrid join. We first sort the outer table on the join column. Then, the outer is joined with the index on the join column of the inner. The inner tuple is represented by its surrogate, equivalent of its physical disk address, which is carried in the index. The partial join result is sorted on the surrogate and then the inner table is accessed sequentially to complete the join result. Local predicate filtering can also be applied before the access of the inner relation through index AND/ORing. We also present efficient methods for skip sequential access and prefetching of logically discontiguous leaf pages of B+-tree indexes.

No abstract