THE STUDY OF 28NM BEOL CU GAP-FILL PROCESS
Yu Bao, Gang Shi, Lin Gao, Yanyan Zhang, Yingming Liu, Peng Tian, Fuchun Xi, Wei Hu, Ying Gao, Zhenhua Cai, Baojun Zhao, Zhigang Yang, Jianghua Leng, Haifeng Zhou, Jingxun Fang Shanghai Huali Microelectronics Corporation, Shanghai 201203, China
ABSTRACT
In this paper, the influence of Copper (Cu) barrier and ed process tuning on step coverage was analyzed. TEM images show relatively thinner barrier can improve the opening CD of a metal line structure hence improve the sidewall coverage of Cu ed. Cu Seed adopts the deposition/re-sputter method to improve the step coverage, and a higher ratio of re-sputter/deposition can increa the thickness of Cu ed on the sidewall. According to the post CMP surface defects scan results, the optimization of barrier Cu ed thickness can significantly reduce the copper void defects density.
INTRODUCTION
Copper has been widely ud in integrated circuit as the interconnect material becau of its low resisti
vity and good electro-migration (EM) resistance. However, As the CD of integrated circuit scaling down, the Cu gap filling became a big challenge. For Cu interconnect formation, Cu barrier ed process is very important, and a robust condition can provide wider process window for subquent electro Cu plating process (ECP). [1-3] Combined with the TEM images of post-Cu barrier/ed deposition and the post-Cu-CMP surface defects data, the influence and mechanism of Cu barrier/ed process tuning on Cu-void performance were propod. EXPERIMENT
Metal line structures were prepared with ULK/MHM scheme. Ta(N) bad barrier and CuMn ed layer were deposited by PVD. To investigate the mechanism of Cu gap filling, the thickness of Cu barrier/ed were measured using TEM images. The gap fill performance was evaluated by surface defects scan at post CMP. RESULTS AND DISCUSSIONS
Barrier Layer
Bilayer Ta(N) bad barrier is deposited by PVD. The whole deposition process is divided into 4 steps, including TaN deposition, Ta deposition, Ar re-sputter, and Ta Flash. TaN is the barrier layer to prevent Cu diffusion. Ta as the wetting layer for Cu ed. Ar re-sputtering can thin down the barrier at the bottom of metal line structures by ion bombardment. The re-sputtered Ta(N) re-deposited onto
the sidewall resulting in bett感恩板报
er step coverage. Figure 1 showed the TEM images of barrier ed with 3 different barrier conditions.
Figure 1. The TEM images of barrier ed with different barrier. (a).Thinner barrier with less Ar re-sputter;
(b).Thinner barrier; (c).Control barrier
The step coverage of Cu barrier ed was shown in Table I. It is difficult to distinguish the boundaries between barrier and Cu ed on sidewall, so we measured the total thickness of秋天壁纸
the barrier and Cu ed. As shown in Figure 2, the thinner barrier improved the opening CD and hence improved the sidewall coverage of Cu ed.
Table I. Normalized step coverage of barrier ed with
Figure 2. Normalized step coverage with different barrier.
(a).Overhang; (b).Sidewall
Cu Seed Layer
Copper ed works as cathode and conductive layer a b
during Cu ECP process. The quality of Cu ed is critical to ECP. Cu Seed formation can be divided into 2 steps, deposition and Cu re-sputter. The Cu re-sputter step can improve the coverage on the sidewall. Figure 3 showed the TEM images of Cu barrier ed with 3 different ed conditions.
Figure 3. The TEM images of barrier ed with different ed. (a).Less deposition ed; (b).Control ed;
(c).More re-sputter ed.
The step coverage of Cu barrier/ed is shown in Table II. The higher the Cu re-sputter/deposition ratio, the higher the Cu ed coverage on表达今天很开心的说说
sidewall.
Table II. Normalized step coverage of barrier ed with
As shown in figure 4, both less deposition and more re-sputter ed can improve sidewall coverage, and less deposition can reduce overhang. So ed with more re-sputter is preferred to gap filling.
Figure 4. Normalized step coverage with different ed.
(a).Overhang; (b).Sidewall
Defect Scan Results
Table III is the design of experiment of Cu barrier/ ed thickness. The gap fill results were evaluated by post Cu CMP void inspection.
Post-CMP defect data in Fig.5 showed thinner barrier is beneficial for ECP 初一祝福语
gap filling due to larger opening CD. However, the thinner ed split can make the Cu void wor significantly.
Figure 5. Normalized Cu void density post Cu-CMP with
various barrier/ed splits
Typical Cu void post CMP was shown in figu氐怎么组词
re 6, missing Cu can be found on the sidewall of recess structure. It ems that continuous ed on the sidewall is more important for gap filling.
Figure 6. Typical Cu void post CMP. (a).Top view;
(b).Cross ction CONCLUSION
In conclusion, Cu-void performance was remarkably improved by the optimization of Cu barrier/ed scheme. Relative thinner barrier combined with thicker continuous ed layer was preferred for subquent Cu ECP. Improved gap fill performance confirmed by post Cu-CMP defect data. Result is consistent with the good step coverage of Cu barrier/ed shown by TEM images.
a b
ACKNOWLEDGEMENTS
The author would like to acknowledge all the 28nm process integration team members of HLMC for the techni怀恋作文
cal discussions, and the failure analysis lab for supporting&日本朝代
nbsp; TEM work.
REFERENCES
[1]Weiye He, Beichao Zhang, Jian Kang, et. al. The
contributions of barrier resputter for BEOL integration. ECS Transactions. 44 (1) 487-492 (2012) [2]Yu Bao, Xuezhen Jing, Jingjing Tan, et. al.
Optimization of Metallization Process for 28-nm-node Low-k /Cu Multilevel Interconnects.
ECS Transactions. 44 (1) 477-480 (2012)
[3]Xuezhen Jing, Jingjing Tan, Jiquan Liu. 32/28NM
BEOL CU GAP-FILL CHALLENGES FOR METAL FILM. CSTIC 2015
