1. Introduction
Copper is the material favored most for integrated circuit (IC) manufacture, owing to its excellent electro migration resistance and low electrical resistance. The integration of copper into an IC manufacturing process can be implemented by using the dual damascene technique, in which chemical mechanical planarization (CMP) has been widely used for planarizing the surface globally and locally[1-4]. Several factors have been considered in the literature, including the effects of numerous variables such as the choices of slurry and pad as well as system facility problems, post-CMP cleaning, and pad conditioning techniques which may affect the CMP process[5-9]. Among the above CMP components, slurry consumables and process variables controlled by equipment are very important parameters in determining removal rate and non-uniformity (within-wafer non-uniformity; WIWNU%). The optimization of CMP process variables by applying the design of experiment (DOE) technique is expected to improve the CMP process ability of new materials and to conserve expense and time, which trial and error techniques may otherwise cause. In this paper, CMP process optimization for bulk copper removal based on alkaline copper slurry was performed on a 300 mm Applied Materials Reflexion LK system. We obtained the optimum process variables by using the DOE method, which can apply the CMP process to the global planarization of multilevel interconnection structures. Optimum process variables have been studied from the viewpoint of removal rate and non-uniformity. Finally, through the above DOE results, we obtain the optimal CMP equipment parameters.
2. Experimental
In this experiment, the CMP process was performed by using an Applied Materials Reflexion LK 300 mm tool with a five-zone polishing head. The polish pad was IC 1010 (DOW Chemical, Co.). The platen/carrier rotary speed was set at a constant value of 97/103 rpm in all main polishing steps based on Semiconductor Manufacturing International Corporation guidelines. The pressure and slurry flow rate were changed from 1.0 to 2.5 psi and from 250 to 400 mL/min for verifying the effect of pressure and slurry flow rate. Dummy wafers were polished for 1 min prior to the main 3 polishing steps. The alkaline copper slurry (namely FA/O copper slurry) applied in this experiment was obtained from the Hebei University of Technology. The slurry solutions included: 4 wt% colloidal silicon with a median particle diameter of 20 nm as the abrasive; 3 wt% polyhydroxy polyamino (FA/O) used as the copper complexing agent and obtained from Tianjin Jingling Mircroelectronics Material Co., LTD; 1 wt% hydrogen peroxide (H
The copper removal rate was measured with a 300 mm copper blanket wafer with 1.2
3. Results and discussion
The DOE technique used for the optimized CMP process is summarized in Table 1.
![]() |
Figure 1 shows the MRR and WIWNU after CMP as a function of various pressures. The polishing conditions were a head/platen speed of 103/97 rpm and a slurry flow rate of 300 ml/min. As shown in Fig. 1, the MRR is increased continuously with the increase of pressure. The average MRR variation between 4640 to 9425 Å/min is observed at applied pressures in the range of 1.0-2.5 psi. Copper removal rates of 5000 Å/min (2.0 psi) or greater are desirable in IC manufacturing where process time and wafer throughput are a major concern. The polishing results obtained from Fig. 1 show that FA/O copper slurry has a higher removal rate of 8553 Å/min at 2.0 psi. WIWNU also increased as the pressure increased, as shown in Fig. 1. The profile of MRR with various pressures indicates that the increases of WIWNU were caused by the low removal rate near the wafer edge and a slightly reduced removal rate around the wafer center. We conclude that as the pressure increases, the removal rate increases and non-uniformity is improved.
The effects of various slurry flow rates on the MRR and WIWNU of a copper blanket wafer are present in Fig. 2. The polishing process was conducted at a pressure of 1.5 psi and a rotational velocity of 103 rpm (polish head) and 97 rpm (platen). As the slurry flow rate is increased, there is no significant improvement in the material removal rate, but there is a slight reduction of the WIWNU and thus an improvement in uniformity. As the flow rate is increased, the MRR profiles presented in Fig. 2 are not changed substantially, and most profiles maintain a shape having a higher removal rate around the center than its edge. The Applied Materials Reflexion LK CMP system applied in this experiment was equipped with 5-zone Contour
Because of the ongoing development of interconnects toward the nanoscale, new interconnect structures are designed to use low-dielectric constant (low-
Apart from achieving high planarization efficiency, reducing the surface roughness of the processed surface also is important for Cu CMP. The surface roughness of copper films has been investigated in the proposed optimum CMP process. The results are presented in Fig. 4, along with the root mean square (RMS) surface roughness and peak/valley (P/V) distances. As shown in Fig. 4, the surface parameters of the copper surface before and after CMP are RMS
4. Conclusions
The process variables of the CMP equipment were apparently dependent on removal rate and non-uniformity. Under the DOE condition, we conclude that as the pressure increases, the removal rate increases and non-uniformity is improved. As the slurry flow rate is increased, there is no significant improvement in the material removal rate, apart from a slight reduction of the WIWNU and thus an improvement of uniformity. We obtained optimized CMP characteristics including a removal rate of over 6452 Å/min and non-uniformity below 4% on a blanket wafer and the step height was reduced by nearly 8000 Å/min in the center of the wafer. On an eight-layer copper patterned wafer, the surface roughness is reduced to 0.225 nm.