High-rate deposition of copper thin films using newly designed high-power magnetron sputtering source

Surface & Coatings Technology 188 189 (2004) 721 727 www.elsevier.com/locate/surfcoat High-rate deposition of copper thin films using newly designed high-power magnetron sputtering source Jin-Hyo Boo a, *, Min Jae Jung b, Heon Kyu Park b, Kyung Hoon Nam b, Jeon G. Han b a CAPST and Department of Chemistry, Sungkyunkwan University, Suwon 440-746, South Korea b CAPST and Department of Metallurgical Engineering, Sungkyunkwan University, Suwon 440-746, South Korea Available online 27 August 2004 Abstract We have deposited copper (Cu) thin films on Si(100) and glass substrates in the growth temperature range between 573 and 753 K using a pulsed DC magnetron sputtering method. Based on the magnetic field simulation, we have designed and constructed a high-power (12010-4 W/m 2 ) unbalanced magnetron sputtering (UBM) source for high-rate deposition. The maximum deposition rate of the newly developed sputtering source under a target power density of 11510-4 W/m 2 we have obtained is 2.8 Am/min. This is five times higher than that using the conventional sputtering method, and the sputtering yield also reached 70% due to low voltage and high-current Cu-accelerated ions. We have also adapted an ion extraction grid between the Cu target and substrate. Although the growth rate was decreased to 2 Am/min, XRD and XPS showed that highly oriented polycrystalline Cu(111) thin films without carbon and oxygen impurities were obtained with lowest electrical resistivity of 2.010 2 AVm at a target power density of 96.710-4 W/m 2, substrate temperature of 723 K, and working pressure of 1.310-1 Pa. During film deposition, moreover, plasma diagnostics was also carried out in situ by optical emission spectroscopy analysis. D 2004 Elsevier B.V. All rights reserved. Keywords: High-rate deposition; Copper thin film; Unbalanced high-power magnetron sputter source; Ion extraction grid; Optical emission spectroscopy 1. Introduction Various metals, such as Ag, Cu, Au, and its alloys have been investigated as possible materials for the replacement of Al and its alloys for electrical interconnects for ULSI applications [1]. Among them, copper (Cu) with its high melting temperature (~1400 vs. ~900 K for Al Cu alloy) is a prime candidate. Moreover, the Cu bulk resistivity is significantly lower than that of aluminum and its alloys (1.710 2 AVm for Cu, while 2.710 2 and 3.010 2 AVm for pure Al and Al Cu alloy), which can translate to reduced signal propagation delay along metal interconnects [2]. Al and its alloys have been commonly used as metallization materials. However, Al suffers from major limitations in electromigration, resistivity, and thermal * Corresponding author. Tel.: +82 31 290 7072; fax: +82 31 290 7075. E-mail address: jhboo@skku.edu (J.-H. Boo). stability. Due to its low electrical resistivity and good electromigration resistance, therefore, Cu is now recognized as the interconnect metal of choice for the next generation of ULSI applications [1 4]. But, there are problems of contamination by C and O and low growth rate in the conventional sputtering method. Therefore, at present, most of the effort is focused on understanding and improving the electromigration performance of aluminum alloys [3,4]. However, a more radical approach that is gaining some attention is the replacement of aluminum altogether by a conductor of significantly higher melting temperature. In order to solve the problems associated with Cu deposition, we have designed and constructed an unbalanced magnetron sputtering (UBM) source with high current and low voltage for high-rate deposition. The conventional methods to fabricate the pure Cu thin films are chemical vapor deposition (CVD), as well as physical vapor deposition (PVD), such as simple sputtering [5 9]. In the case of CVD however, normally there are some 0257-8972/$ - see front matter D 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.surfcoat.2004.07.005

722 J.-H. Boo et al. / Surface & Coatings Technology 188 189 (2004) 721 727 impurities in the films due to the incorporation of carbon and oxygen atoms from precursor and/or residual gas. Furthermore, the simple sputtering becomes insufficient because of its broad angular distribution of the atom flux that leads to pinch-off near the opening of the trench and void formation in the films during the trench filling [10]. Recently, the RF and DC magnetron sputtering with balanced or unbalanced magnets were proposed to solve the aforementioned issues [11 13]. Among them, an unbalanced magnetron (UBM) sputtering method by pulsed DC was our focus to improve the film crystallinity together with mechanical and protective properties of the Cu thin films, which are grown at relatively low temperature. Because pulsed DC magnetron sputtering has many advantages, such as high growth rate, low temperature deposition, and good reproducibility [12], in this study, we intended to make high-quality Cu thin films at a temperature as low as 573 K using this technique. Moreover, we adapted an ion extraction grid on the newly designed unbalanced magnetron sputter source to further improve film properties. 2. Experimental Experiments were carried out using a homemade sputtering system with newly developed high-rate and high-power magnetron sputter source, as schematically shown in Fig. 1(a). We designed and constructed a circular planar unbalanced magnetron sputter source (see Fig. 1(b)) with a target diameter of 0.1 m, electromagnets, and high target power density capability; that is, the sputter source can operate under the extreme condition, such as high vacuum (less than 10 1 Pa), high current (20 12010 4 ma/m 2 ), and low voltage (100 1000 ev). Before construction, a magnetic field simulation had been in first carried out using a computer program in order to improve the sputtering yield and the growth rate. Based on the simulation, we built the high-power magnetron sputter source as shown in Fig. 1(b) that has unbalanced magnetrons on both inside and outside of the chamber. The most homogeneous magnetic filed distribution was observed when we applied the magnetic filed of 0.03 Tesla into both the inner and the outer magnetic coils. It was noticeable that in the case of balanced magnetic coil, inhomogeneous magnetic field has been induced by the result of our computer simulation. However, when we took an balanced magnetic coil system with I inner coil /I outer coil=1:2, we obtained the best homogeneous magnetic field, suggesting that the sputtering yield together with growth rate can be increased as much as a factor of 2. This is in good agreement with the previous report published by Kadlec and Musil [12]. In this study, furthermore, we found that with adapting an ion extraction grid located between Cu target and substrate, the deposition rate and sputtering yield can additionally increase by a factor of two or three times. Fig. 1(c) shows the principle of Cu ion extraction grid taken in our system. The enhancement of the deposition rate and plasma density by attachment of the Cu ion extraction grid can be explained as the following. The grid with a negative electric potential ( 650 V) will reflect electrons and restrict them between the grid and the target surface and thereby increasing electron density. Such enhancement of electron density will then increase the collision probability of electrons with neutrals including Ar and copper vapor and promote further ionization of Ar and copper vapor. The sputtering rate of the target surface will then be improved by the higher flux Ar ion bombardment. Consequently, more copper vapor becomes ionized. Such combined effect of high vapor flux and ion extraction by grid will contribute the enhancement of deposition rate of film as well as electromigration, such as surface mobility, resulting in low electrical resistivity. The maximum deposition rate and sputtering yield obtained in this study using the newly designed sputter source obtained without and with the ion extraction grid are 2.8 Am/min and 50%, and 2.0 Am/min and 70%, respectively. This is three or five times higher than those using conventional sputtering methods. We used a pulsed DC of 3.333-kHz frequency and 50% duty ratio to improve the crystallinity and grain size of the Cu thin film grown at a low-temperature growth of below 753 K. Si(100) and Corning 1737 glass were used as substrates, which were heated by either a heater or plasma in the temperature range between 573 and 753 K. The substrate temperature was measured using an optical pyrometer. The general growth conditions of the Cu thin films are 2.710 2 6.710 1 Pa of working pressure, 573 753 K of deposition temperature, and 650 V of bias voltage on the Cu ion extraction grid. The distance between the target and the grid was maintained at 0.05 m, while that between the substrate and the target at 0.1 m. The as-grown Cu thin films were characterized with XRD and XPS, and the film thickness was obtained by an a-step profiler and cross-sectional SEM analysis. The electrical resistivity of the as-grown Cu thin films was also measured using a four-point probe method. During sputtering, the characteristics of the Cu sputter source were also analyzed with an in situ Langmuir probe method, and the optical emission spectroscopy (OES) to study the gas-phase reaction in the plasma. 3. Results and discussion Fig. 2(a) shows the current voltage (I V) characteristics of the UBM source developed in this work with a Cu target and a pulsed DC power supply. As can be seen, the discharge voltage increases with the decrease of Ar partial pressure, and it can be sustained down to 2.710 2 Pa even at a higher discharge current of 15 A. This high-power discharge at low

J.-H. Boo et al. / Surface & Coatings Technology 188 189 (2004) 721 727 723 Fig. 1. (a) Schematic diagram of the high-rate sputtering system. (b) Newly developed unbalanced high-power magnetron sputter source for Cu thin film deposition: (1) target, (2) inner magnetic coil, (3) outer magnetic coil, (4) cooling jacket, (5) insulator, (6) centered steel bar, and (7) outer steel plate. (c) The principle of copper ion extraction using a negatively biased grid.

724 J.-H. Boo et al. / Surface & Coatings Technology 188 189 (2004) 721 727 Fig. 2. (a) Current voltage (I V) characteristics of the UBM source developed in this work. (b) The changes of target temperature with target power density. pressure is due to the high strength and density of magnetic field on the target surface. Therefore, unlike conventional magnetrons, this magnetron can operate at low pressures but still maintain high rates. The temperature variation of target surface at 5 min of operation, measured by an Infrared (IR) optical pyrometer, as a function of target power densities is illustrated in Fig. 2(b). Up to the target power density of 8010 4 W/m 2, the temperature of the target surface was maintained at about 693 K, but the temperature increased rapidly at higher target power densities, indicating that our UBM source can be stably operated at target power density of lower than 8010 4 W/m 2. As we understand, the target power density and target temperature are very important to obtain high-rate deposition, and a maximum deposition rate of the film can be limited by the cooling of the sputtered target. In this study, a mixture of ethylene glycol (HOC 2 H 4 OH) and deionized water was used as the coolant of the target in a volume ratio that is 1:1, and it was kept at 268 K by a cooling system. Fig. 3(a) shows a variation of deposition rate of Cu films with increasing target power density under our deposition condition. The maximum deposition rate was reached to about 2.8 A/min at 11510 4 W/m 2 of target power density. At the target power density of 8010 4 W/m 2, which corresponds to a stable operating without overheating, a deposition rate of about 2.0 A/min was achieved. This deposition rate is at about five times that using ordinary magnetrons. Target erosion occurred along the maximum strength and density of the magnetic field on the target surface, and a target utilization of 25 30% was achieved. When an ion extraction grid was adapted between the Cu target and substrate however, the growth rate rapidly decreased due to the shielding effect. OES is a powerful tool for controlling the sputtering and deposition process and for optimizing both the Cu film layers properties. In this study, we used OES for qualitative in situ plasma diagnostics. Fig. 3(b) and (c) show the optical emission spectra of Cu neutrals and ions obtained both

J.-H. Boo et al. / Surface & Coatings Technology 188 189 (2004) 721 727 725 Fig. 3. (a) The variation of Cu film deposition rate grown either without (closed circle) or with (open circle) the ion extraction grid as a function of target power density. (b) and (c) The optical emission spectra of Cu ion obtained by both without (b) and with (c) the ion extraction grid under the same deposition condition (i.e., working pressure: 1.310 1 Pa; target power density: 96.710 4 W/m 2 ). without (3b) or with (3c) ion extraction grid under the same deposition conditions (i.e., working pressure: 1.310 1 Pa; target power density: 96.710 4 W/m 2 ). Three different OE spectra were observed at 353.4, 359.7, and 369.1 nm. By comparing these experimental data with the possible theoretical transition, we identified these OE peaks as following: the first and second peaks at 353.4 and 359.7 nm are attributed to the Cu neutrals transition, whereas the third peak at 369.1 nm mainly to a Cu + ion transition. In Fig. 3(b) and (c), we can find that the emission peak intensities of the Cu neutrals and ions with ion extraction grid exhibit approximately three times larger than those of without grid. This indicates that the negatively biased grid may play an important role in enhancing the plasma density of Cu ions as well as Cu neutrals from intensive Ar + ion sputtering and thereby lading to Cu thin films with high electron mobility. Fig. 4(a) shows the typical XRD pattern of a Cu thin film deposited on glass substrate at a target power density of 96.710 4 W/m 2, a substrate temperature of 723 K, and a working pressure of 1.310-1 Pa without Cu ion extraction grid. Three diffraction peaks of Cu(111), Cu(200), and Cu(220) can be seen in the Fig. 4(b), signifying polycrystallinity of the Cu film. The electrical resistivity obtained from this film is about 4.010 2 AVm. To investigate the effect of the ion extraction grid on the film structure, we have also deposited Cu films on glass substrates under the same deposition conditions as for Fig. 4(a). As much higher quality polycrystalline Cu film with lower electrical resistivity, low impurity and a low growth rate than that of Fig. 4(a) was obtained. In Fig. 4(b), a strong diffraction peak attributed to the Cu(111) phase is observed at 2h=438, as well as some small diffraction peaks of Cu(200) and Cu(220), suggesting a highly oriented poly-

726 J.-H. Boo et al. / Surface & Coatings Technology 188 189 (2004) 721 727 Fig. 4. XRD patterns of Cu thin films grown on glass substrates at target power density of 96.7 W/cm 2, substrate temperature of 723 K, working pressure of 1.310 1 Pa without Cu ion extraction grid (a) and with Cu ion extraction grid (b). (c) A typical X-ray photoelectron (XP) survey spectrum and high-resolution XP spectrum (insert) of Cu 2p obtained from the same film as (b). crystalline film in the [111] direction. The obtained Cu film with approximately 3.5 Am thickness exhibits (111) preferred orientation and a low electrical resistivity of about 2.010 2 AVm that is comparable with the electrical resistivity (1.710 2 AVm ) of bulk Cu. This suggests that the ion extraction grid has a strong effect on the structural and electrical properties of the Cu films. To check whether the film crystallinity as well as impurity, such as carbon and oxygen, can influence to the electrical resistivity or not, we carried out X-ray photoelectron (XP) spectra measurements. Fig. 4(c) shows a typical X-ray photoelectron (XP) survey spectrum and highresolution XP spectrum (insert) of Cu 2p obtained from the same film as Fig. 4(b). From the survey spectrum, we can see only copper peaks, indicating that there are no carbon and oxygen contamination induced from either target poisoning or incorporation of residual gas. Furthermore, the high-resolution spectrum of copper 2p shows a sharp peak at about 932.7 ev, signifying the formation of a pure metallic copper film (see insert). Therefore, we can obtain the best Cu film with good crystallinity and conductivity with the ion extraction grid.

J.-H. Boo et al. / Surface & Coatings Technology 188 189 (2004) 721 727 727 4. Conclusions New unbalanced magnetron sputtering source was designed and constructed for improving growth rate and target yield. Due to the homogeneous magnetron field of 0.03 Tesla, we were able to operate the target power density up to maximum 11510 4 W/m 2 using a 20-kW (pulsed) DC power supply. With the new magnetron source, Cu thin films with no C and O contamination were deposited. XPS spectra of Cu films exhibited high-quality metallic Cu thin films with no impurity, resulting in good electrical conductivity. Using the ion extraction grid, moreover, we deposited a Cu thin film with (111) preferred orientation and low electrical resistivity (2.010 2 AVm). The maximum deposition rate and sputtering yield of the newly developed sputtering source under a target power density of 11510 4 W/m 2 in Ar pressure of 1.310-1 Pa were 2.8 A/min and 70%, respectively. This is five times higher than that of the conventional sputtering method. During film deposition, in situ plasma diagnostics were also carried out using optical emission spectroscopy, and ions and neutrals of both Ar and Cu were observed. Higher intensities of the ions and neutrals with the grid were attributed to the high-quality Cu film formation. Acknowledgements Support of this research by the Center for Advanced Plasma Surface Technology of the Sungkyunkwan University and by the Ministry of Science and Technology of Korea (Project No. M10214000278-02B1500-04211) is gratefully acknowledgment. References [1] G.H. Takaoka, J. Ishikawa, T. Takagi, J. Vac. Sci. Technol., A, Vac. Surf. Films 8 (1990) 840; G.H. Takaoka, K. Fujita, J. Ishikawa, T. Takagi, Nucl. Instrum. Methods, B 37 38 (1989) 882. [2] S.P. Murarka, S.W. Hymes, Crit. Rev. Solid State Mater. Sci. 20 (2) (1995) 87. [3] C.C. Lee, E.S. Machlin, H. Rathore, J. Appl. Phys. 71 (1992) 5877; J. Li, J.W. Mayer, J. Appl. Phys. 70 (1991) 2820. [4] P. Bai, G.R. Yang, L. You, T.M. Lu, D.B. Knorr, J. Mater. Res. 5 (1990) 989; J. Onuki, Y. Koubuchi, M. Suwa, D. Gardner, H. Suzuki, E. Minowa, IEEE Trans. Electron Devices 39 (1992) 1322. [5] N. Awaya, Y. Arita, Jpn. J. Appl. Phys. 30 (1991) 1813. [6] A.S. Jain, K.-M. Chi, T.T. Kodas, M.J. Hamden-Smith, J.D. Farr, M.F. Paffett, Chem. Mater. 3 (1991) 995. [7] A.E. Kaloyeros, M.A. Fury, Mater. Res. Soc. Bull. 18 (1993) 22; S. Cohen, M. Liehr, S. Kasi, Appl. Phys. Lett. 60 (1992) 50. [8] S.M. Rossnagel, J. Vac. Sci. Technol., A, Vac. Surf. Films 13 (1995) 125; W. Posadowski, Surf. Coat. Technol. 49 (1991) 290. [9] W. Eang, J. Foster, A.E. Wendt, J.H. Booske, T. Onuoha, P.W. Sandstrom, H. Liu, S.S. Gearhart, N. Hershkowita, Appl. Phys. Lett. 71 (1997) 1622. [10] S.M. Rossnagel, J. Hopwood, J. Vac. Sci. Technol., B 12 (1) (1994) 449. [11] H. Jiang, T.J. Klemmer, J.A. Barnard, E.A. Payzant, J. Vac. Sci. Technol., A, Vac. Surf. Films 16 (1998) 3376; H. Jiang, T.J. Klemmer, J.A. Barnard, W.D. Doyle, E.A. Payzant, Thin Solid Films 315 (1998) 13. [12] S. Kadlec, J. Musil, Vacuum 47 (1996) 307; J. Musil, A. Rajsky, A.J. Bell, J. Matous, M. Cepera, J. Zeman, J. Vac. Sci. Technol., A, Vac. Surf. Films 14 (1996) 2187. [13] S. Ghosh, K. Hong, C. Lee, Mater. Sci. Eng., B, Solid-State Mater. Adv. Technol. 96 (2002) 53; L.H. Qian, Q.H. Lu, W.J. Kong, K. Lu, Scr. Mater. 50 (2004) 1407.