Plasma-Based Copper Etch Process and Reliability

Abstract

Copper (Cu) has been used to replace aluminum as the alternative interconnect material because of its low resistivity, lack of hillock formation, and high resistance to eletromigration. When width of the interconnect lines continues to decrease, the copper film structure and edge profile have a more significant impact on the interconnect reliability. Previously, a plasma-based Cu etch process was developed.[1, 2] In this paper, the authors investigated the effect of etching and deposition conditions on reliability of the copper line interconnect. A 200 nm Cu film was deposited on a thin layer of TiW that was previously deposited on glass. Both films were deposited by sputtering in Ar using 13.56 MHz power supply. The grain size of the copper film was calculated from XRD peaks. The film resistivity was determined from the current-voltage (I-V) curve. After sputter deposition, the Cu film was patterned into different line widths using a positive photoresist. Then, the patterned Cu film was exposed to the plasma composed of Cl2, Ar or/and CF4 feed gases in a reactive ion etching (RIE) reactor with a parallel electrode arrangement. Subsequently, the sample was submerged in a dilute HCl solution for 10 seconds to remove the CuClx products. Scanning electron microscope (SEM) images were taken to examine the edge roughness of the etched lines. For the step of sputtering deposition of Cu film, both deposition rate and film quality were of concern. Chamber pressures of 3, 5, and 10 mTorr and RF powers of 50, 70, and 90 W were used to find such a point that allows both a high deposition rate and a small grain size. A high deposition rate is preferred to increase the production rate, and small grain size is desirable for the reduction of the edge roughness during the etching reaction process. Comparison of Cu films obtained from the 9 conditions showed that the pressure of 10 mTorr and RF power of 90 W could deposit a Cu film at 32.3 Å/min with a grain size of 178 Å, as shown in Figure 1. Although the grain size was slightly larger compared to the smallest grain size obtained at 161 Å, under the power of 50 W and the same pressure of 10 mTorr, the deposition rate at 90 W was notably higher when compared to the deposition rate of 19.2 Å at 50 W. Higher deposition rate at high power could be explained by the high bombardment energy, which resulted in the higher sputtering yield of the target. The small grain size at the low-power, high-pressure deposition condition could be attributed to the lower transfer energy from the target and the low surface migration rate of the adatoms. In addition, at 10 mTorr, the resistivity of the Cu film deposited at 90 W is significantly lower than that deposited at 50 W, as shown in Figure 2. The resistivity difference can be contributed by the grain size difference. When electrons are transferred in the small grain film, they are prone to scattering at grain boundaries and therefore, the film has a large resistivity. For the etching process, a high etching rate and a smooth pattern edge are important. For the conversion of Cu to CuClx in the plasma-copper reaction, experiments were carried out at different pressures, i.e., 40, 80, and 120 mTorr, separately, powers, i.e., 300, 500, and 600 W, and feed gas streams. Results show that when the feed gas is composed of Cl2/CF4 10/5 sccm, at 600 W and 40 mTorr, the highest Cu-to-CuClx conversion rate was obtained. Under this condition, the highest self-bias voltage, i.e., VDC, was obtained. Figure 3 shows the 20k magnification SEM image of the final Cu line edge, which has a small roughness. Cu lines of various widths were stressed at 7 V to investigate the electromigration phenomena. Times required for the breakage of the lines were recorded and compared. The results showed that the thinner lines broke in a shorter period of time. The 10, 20, and 40 µm wide Cu lines broke in 252, 337, and 442 seconds respectively. The authors acknowledge the financial support of this work through NSF projects 1633580 and 1633500. Lee and Y. Kuo, JECS. 148 (9), G524-529 (2001).Kuo and S. Lee, Appl. Phys. Lett. 78, 1002 (2001). Figure 1