Abstract:We have studied electromiyation (EM) and stress-induced voiding (SV) behaviors based on our 90nm-node Cuilow-k interconnect processes, and demonstrated successful improvement of the interconnect reliability. In EM study wide bimodal failure distribution was found only in the particular EM test structure. We identified that it caused by the lack of wettability between Cu and the barrier metal in the vias, and demonstrated that the ,optimization of the barrier metal thickness could suppress it. In SV behavior, w… Show more
“…4 TaN currently used as a diffusion barrier can be replaced with Ta/TaN to improve the adhesion strength of Cu, and thus enhance the electromigration lifetime. [6][7][8] However, continued aggressive reduction in transistor dimensions requires a new generation of underlayers with low resistivity and better adhesion property with Cu for low contact resistance and long electromigration lifetime.…”
The main issue of Cu metallization is the electromigration of Cu through the interface between Cu and the barrier or capping layer. To improve electromigration resistance at the Cu and barrier metal interface, insertion of a glue layer which enhances the adhesion of Cu onto the under layer may be effective. The wettability of Cu on Ru and Ta glue layers was evaluated as the index of Cu adhesion strength onto glue layers. The wetting angle of Cu (43°) on a Ru substrate was three times lower than that of Cu (123°) on a Ta substrate after annealing. Lower wetting angle of Cu on a Ru substrate indicates a good adhesion property between Cu and Ru and may imply a high electromigration resistance. The better Cu wettability of Ru compared to Ta can be explained by the concept of lattice misfit. A Ru(002) plane has lower lattice misfit, which suggests lower interface energy, and enhanced the adhesion of Cu onto Ru. However, the Ru film showed poor Cu diffusion barrier properties, which suggests Ru should be used as a glue layer in combination with another barrier layer.
“…4 TaN currently used as a diffusion barrier can be replaced with Ta/TaN to improve the adhesion strength of Cu, and thus enhance the electromigration lifetime. [6][7][8] However, continued aggressive reduction in transistor dimensions requires a new generation of underlayers with low resistivity and better adhesion property with Cu for low contact resistance and long electromigration lifetime.…”
The main issue of Cu metallization is the electromigration of Cu through the interface between Cu and the barrier or capping layer. To improve electromigration resistance at the Cu and barrier metal interface, insertion of a glue layer which enhances the adhesion of Cu onto the under layer may be effective. The wettability of Cu on Ru and Ta glue layers was evaluated as the index of Cu adhesion strength onto glue layers. The wetting angle of Cu (43°) on a Ru substrate was three times lower than that of Cu (123°) on a Ta substrate after annealing. Lower wetting angle of Cu on a Ru substrate indicates a good adhesion property between Cu and Ru and may imply a high electromigration resistance. The better Cu wettability of Ru compared to Ta can be explained by the concept of lattice misfit. A Ru(002) plane has lower lattice misfit, which suggests lower interface energy, and enhanced the adhesion of Cu onto Ru. However, the Ru film showed poor Cu diffusion barrier properties, which suggests Ru should be used as a glue layer in combination with another barrier layer.
“…This success of the simpler and hence, manufacturable integration with less materials and interfaces is attributed to the capability of modern dry etching tools which have controllability and uniformity across the wafer enough to support the pattern definition in the monolithic damascene scheme. Another example to show the strong dependency on the tool/material/process capability and readiness is the across-the-wafer non-uniformity of modern CMP tools which so far seem unable to support the Hard Mask Retention scheme for the use of ULK at the trench level for the 65 nm node BEOL technologies in the manufacturing phase, although a bunch of damascene schemes have been reported [18][19][20][21][22][23][24].. Thus, discussions on possibility of implementation of ULK in the 45 nm node BEOL must be made on certain assumptions about the tool capability/readiness and limitation for manufacturing [25][26][27].…”
Section: Technical Challenges To the 45 Nm Beol Processmentioning
This paper discusses low-k/copper integration schemes which has been in production in the 90 nm node, have been developed in the 65 nm node, and should be taken in the 45 nm node. While our baseline 65 nm BEOL process has been developed by extension and simple shrinkage of our PECVD SiCOH integration which has been in production in the 90 nm node with our SiCOH film having k=3.0, the 65 nm SiCOH integration has two other options to go to extend to lower capacitance. One is to add porosity to become ultra low-k (ULK). The other is to stay with low-k SiCOH, which is modified to have a "lower-k". The effective k-value attained with the lower-k (k=2.8) SiCOH processed in the "Direct CMP" scheme is very close to that with an ULK (k=2.5) SiCOH film built with the "Hard Mask Retention" scheme. This paper first describes consideration of these two damascene schemes, whose comparison leads to the conclusion that the lower-k SiCOH integration can have more advantages in terms of process simplicity and extendibility of our 90 nm scheme under certain assumptions. Then describing the k=2.8 SiCOH film development and its successful integration, damascene schemes for 45nm nodes are discussed based on our learning from development of the lower-k 65nm scheme. Capability of modern dry etchers to define the finer patterns, nonuniformity of CMP, and susceptibility to plasma and mechanical strength and adhesion of ULK are discussed as factors to hamper the applicability of ULK.
“…Before developing our discussion, we have to comment on the merit of lowering VM potential. Various methods of controlling source line potential to reduce leakage current have been reported so far (International Technology Roadmap for semiconductors, 2001; Matsumoto et al, 2003;Nii et al, 1998;Oh et al, 2003;Osada et al, 2001Osada et al, , 2003Sze, 1981). Lowering the V DD source line has the advantage of reducing the sub threshold leakage current of the PMOS load transistor.…”
In sub-100 nm generation, gate-tunneling leakage current increases and dominates the total standby leakage current of LSIs based on decreasing gate-oxide thickness. Showing that the gate leakage current is effectively reduced by lowering the gate voltage, we propose a local DC level control (LDLC) for static random access memory (SRAM) cell arrays and an automatic gate leakage suppression driver (AGLSD) for peripheral circuits. We designed and analyzed a 32 kb 1-port SRAM using 45 nm CMOS technology. The six-transistor SRAM cell size is 1.25 µm 2. Evaluation shows that the standby current of 32 kb SRAM is 1.2 µA at 1.2 V and room temperature. It was reduced to 7.5% of the conventional SRAM.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.