Wear 257 (2004) 785–789
Chemical mechanical polishing by colloidal silica-based slurry for micro-scratch reduction Yoomin Ahn a,∗ , Joon-Yong Yoon a , Chang-Wook Baek b , Yong-Kweon Kim b a b
Department of Mechanical Engineering, Hanyang University, 1271 Sa-1-dong, Sangnok-gu, Ansan, Kyeonggi-do 426-791, South Korea School of Electrical Engineering and Computer Science, Seoul National University, Kwanak P.O. Box 34, Seoul 151-600, South Korea Received 4 August 2003; received in revised form 17 March 2004; accepted 17 March 2004
Abstract The chemical mechanical polishing (CMP) of aluminum and photoresist using colloidal silica-based slurry was investigated. The effects of varying slurry pH, silica concentration and oxidizer concentration on surface roughness and removal rate were investigated in order to determine the optimum conditions for those parameters. Using these optimum conditions silica-based CMP was compared with conventional CMP, which uses an alumina-based slurry. The results of the CMP of the aluminum with the colloidal silica-based slurry were good, but the CMP of the photoresist were not. The colloidal-based silica slurry produced a desirable fine Al surface with few micro-scratches, which is similar to what is produced by CMP using a filtered alumina-based slurry, but produced a photoresist surface with many micro-scratches. © 2004 Published by Elsevier B.V. Keywords: Chemical mechanical polishing; Micro-scratch; Colloidal silica-based slurry; Micro electro mechanical systems
1. Introduction Chemical mechanical polishing (CMP) is one of the important processes in the fabrication of integrated circuits (IC) [1,2]. Currently, the use of CMP in micro electro mechanical systems (MEMS) is increasing [3]. Micro-scratches usually occur on a surface when the metal and dielectric thin films are polished by CMP. The micro-scratches diminish the reliability and endurance as well as the reflectivity of the surface. For example, when the micro-scratches are on the film surface, a short circuit caused by electromigration can occur and the corrosion resistance can be lowered in the metal film, and the current leakage can occur and the breakdown strength can reduce in the polymer dielectric film [4]. Hence, it is necessary to reduce micro-scratches resulting from in CMP process. It has been reported that the major cause of micro-scratches are the very large particles in the slurry [5]. According to a dielectric oxide CMP experiment by Basim et al. [6], the frequency of micro-scratches increased as the size of the largest abrasive particles was enlarged and the number of it increased. Kallingal et al. [7] found that micro-scratches increased as the hardness of the polishing pad increased in an Al CMP. Zhong et al. [8] observed that the density ∗ Corresponding author. Fax: +82-31-406-5550. E-mail address:
[email protected] (Y. Ahn).
0043-1648/$ – see front matter © 2004 Published by Elsevier B.V. doi:10.1016/j.wear.2004.03.020
of the micro-scratches increased with increasing polishing pressure and the size of the micro-scratch increased with enlarging abrasive particles. Mechanical abrasion by the sliding indentation of abrasive particles is believed to be as the main mechanism causing the micro-scratch [8]. It would seem that using a softer polishing pad or reducing the polishing pressure would be good way for reduce micro-scratching. However, using a soft pad generally causes the planarity to deteriorate, and lowering the polishing pressure results in a decrease in the removal rate [9]. In order to reduce micro-scratching, Hara et al. [10] used an abrasive material that induced less mechanical indentation abrasion. Kondo et al. [11] used a slurry that contained no abrasive particles. They reported good results. The goal of this study was to reduce micro-scratching resulting during CMP. We investigated the effect of altering three parameters of the CMP process-slurry pH, silica concentration and oxidizer concentration. The tested materials were aluminum metal that is used for the electric wiring of IC and the organic polymer, photoresist (PR), that is used as a sacrificial layer in MEMS. Al and PR were selected because they are easily scratched. We examined whether micro-scratching during CMP could be reduced by using colloidal silica, which has a lower hardness and smaller mean diameter than alumina.
786
Y. Ahn et al. / Wear 257 (2004) 785–789
2. Experimental
3.1. Aluminum CMP
(1)
An aluminum oxide layer forms on the aluminum surface. The brittle oxide is removed by the abrasive particle in the slurry and is solved in the slurry by the chemical etching. During the CMP process, the formation and removal of the oxide layer is repeated. The material removal rate is greater when the oxidized aluminum is fractured than when the metal aluminum is grinded by the abrasive particle.
Roughness, Rv(nm)
100
100 80
80 60
60
40
40 20
20
0 2
4
6
8
0 12
10
pH Fig. 1. Roughness and removal rate of Al films as a function of slurry pH.
Fig. 1 plots the micro-scratch size (Rv) and removal rate (RR) dependence as a function of the pH of the abrasive slurry. At lower pH values, small Rv values and large removal rates were observed. The Rv increased and the removal rate decreased as the pH increased. It was thought that the chemical reaction was activated at lower pH values, but the reaction became deactivated as the pH value increased [13]. In the deactivated chemical reaction state, oxidization and etching became negligible. Therefore, the removal rate decreased, and the micro-scratches increased since the metal surface under very thin oxide layer was mechanically abraded. In Fig. 1, the micro-scratch size (Rv) was greater at pH 6–11 than at pH 2–4. It was thought that because the isoelectric point of the zeta potential of the colloidal silica was close to pH 3, the abrasive particles which had no potential were most easily agglomerated at pH ∼3. As the size of agglomerated particles increased, the micro-scratch size increased. It was observed that the isoelectric point of fumed silica was about pH 3 [14]. For a better understanding, further study, for example, about the zeta potential of colloidal silica, is needed. Fig. 2 shows the effect of H2 O2 concentration on the Al CMP. The micro-scratch size (Rv) was small from 1 to 120
140 Rv : pH4, SiO2(2wt%) Rv : pH2, SiO2(5wt%) RR : pH2, SiO2(5wt%)
120 100
100 80
80 60 60 40
40
20
20 0
0
2
4
6
8
10
0
Removal rate, RR (nm/min)
Al CMP using a colloidal silica-based slurry was performed by varying the slurry pH, oxidizer H2 O2 concentration, and abrasive concentration in the slurry. The polishing time was 4 min. During the CMP, the chemical reaction on aluminum surface is believed to be as follows [13]:
120
Roughness, Rv (nm)
3. Results and discussion
Rv : H2O2(3vol%), SiO2(2wt%) Rv : H2O2(10vol%), SiO2(5wt%) RR : H2O2(10vol%), SiO2(5wt%)
Removal rate, RR (nm/min)
The goal of the Al and PR CMP experiments by colloidal silica-based slurry was to obtain a scratch-free surface. It is thought that if blunt particles of colloidal silica are used instead of sharp particles of fumed silica, micro-scratching could be reduced. Al of 1 m thickness was sputtered on 4 in. silicon wafers at 480 W and 6 mTorr Ar. PR films were spin-coated with AZP4620 PR on 4 in. Si wafers with a thickness of 7 m and then baked sequentially at 95 ◦ C for 30 min, 120 ◦ C for 30 min, and 180 ◦ C for 1 h. Test samples of Al and PR were segmented into 20 mm × 20 mm size by a dicing saw. A Motopol 2000 polishing system from Buehler Co. was modified for the CMP experiments. The slurry was made by adding colloidal silica abrasive (average ∼20 nm diameter, Ace Hitech) to de-ionized water. H3 PO4 and KOH were added to adjust the pH. In the Al CMP, H2 O2 was used as an oxidizer. In all experiments, the polishing pressure was 30 kPa, the linear polishing velocity was 15 m/min, and the slurry flow rate was 12 ml/min. Polytex Supreme pad from Rodel was used. Micro-scratch and polishing removal rates were measured after CMP. Since the size distribution of the abrasive is approximately normally distributed [12], the maximum size of the micro-scratches could increase as the frequency of micro-scratching increases. In this work, the parameter Rv (the maximum depth of profile) was used to evaluate the size of the micro-scratch. The measurement was carried out under a sampling length of 10 mm, a cut off length of 0.08 mm, and a scanning speed of 25 m/s by Dektak (Veeco Instruments). The removal rate of Al was measured by a four-point probe (Chang Min Tech) and PR removal rate was measured under a wave length of 480 nm and a reflective index of 1.63 by a NanoSpec (Nanometrics).
2Al + 3H2 O2 → Al2 O3 + 3H2 O → 2Al(OH)3
120
140
H2O2 (vol%) Fig. 2. Roughness and removal rate of Al films as a function of H2 O2 concentration.
Y. Ahn et al. / Wear 257 (2004) 785–789
100
80
80 60 60 40
40
20
20 0
0
2
4
6
8
10
12
14
16
18
0
Rv : SiO2(4wt%) Rv : SiO2(10wt%) RR : SiO2(10wt%)
150 125
150 125
100
100
75
75
50
50
25
25
0
2
4
6
8
10
12
Removal rate,RR (nm/min)
100
175
Removal rate, RR (nm/min)
120
Roughness, Rv (nm)
120
Rv : pH4, H2O2(7vol%) Rv : pH2, H2O2(3vol%) RR : pH2, H2O2(3vol%)
Roughness, Rv (nm)
140
787
0
SiO2 (wt%)
pH
Fig. 3. Roughness and removal rate of Al films as a function of SiO2 concentration.
Fig. 4. Roughness and removal rate of PR films as a function of slurry pH.
175
Roughness, Rv(nm)
125
It was reported that the chemical reaction, such as etching, in PR CMP was not active as it was in Al CMP [14].
150 125
100
100
75
75
50
50
25
25
0
3.2. Photoresist CMP
Rv : pH2 Rv : pH4 RR : pH4
150
0
2
4
6
8
10
12
14
16
0 18
Removal rate, RR (nm/min)
3 vol.% H2 O2 and increased with increasing H2 O2 concentration. The removal rate was maximum at a concentration of 1 vol.% and decreased as the concentration was increased above 1 vol.%. It was thought that from 1 to 3 vol.% H2 O2 , the formation rate and the removal rate of the oxide layer were almost equal, so that the removal rate optimally increased and the polished surface was obtained with few micro-scratches. As the oxidizer concentration increases, the passive film of aluminum oxide becomes dense so that the diffusion of oxygen into the aluminum surface beneath the oxide film significantly decreases. Hence, it seems that the formation rate of the oxide became smaller than its removal rate by abrasive grinding. The brittle fracture of the oxide occurred less and the removal rate of CMP decreased. The ductile sub-metal rather than the oxide layer was grinded by abrasive particles so that the rough surface with many micro-scratches formed. In addition, when the slurry without the oxidizer was used, the micro-scratch size (Rv) was very large because the polishing was performed only by mechanical abrasion. The effect of the abrasive concentration on the microscratch size and removal rate are shown in Fig. 3. With a more abrasive in the slurry, the micro-scratch size (Rv) increased or remained constant. As the abrasive concentration increased, the removal rate increased, but decreased after 11 wt.%. It was thought that if more abrasive is used, the frequency of the wear from mechanical abrasion increases. Therefore, the removal rate and the micro-scratch size increased. However, in some cases, the removal rate did not change and the micro-scratch size decreased. The reason may be that the chemical reaction, such as the formation of oxide layer, could have varied as the abrasive concentration increased. For a clearer understanding of that, further study is required.
Therefore, PR CMP was carried out only with variations of the slurry pH and the abrasive concentration. The polishing time was 5 min. Fig. 4 shows the removal rate and micro-scratch size as a function of the slurry pH. As the pH value increased, the micro-scratch size and removal rate increased. It is thought that the surface of PR becomes more chemically active at higher pH values so that the bonding force of atoms and molecules in the surface becomes weaker. Hence, the weakened surface became rougher by abrasion with the abrasive and more surface material was removed. Commercial colloidal silica slurry has pH of approximately 9. The reason may be that at the pH ∼9, the silica particles have large potential with the same sign so they repel each other without agglomeration. Hence, it is expected that when the alkaline solution is used rather than acid one, the PR surface will be fine since the abrasion occurs with small abrasive particles during CMP. Nevertheless, the experimental results were different from the expectation. This might be due to the fact that the zeta potential of the particles was not as significant as chemical activation on the surface. The dependence of CMP on the abrasive concentration is shown in Fig. 5. Generally, the surface became rougher and the removal rate increased as the abrasive concentration was increased. The reason seems to be that abrasion was more
SiO2 (wt%) Fig. 5. Roughness and removal rate of PR films as a function of SiO2 concentration.
788
Y. Ahn et al. / Wear 257 (2004) 785–789
severe when the concentration of the abrasive particles was increased. 3.3. Evaluation of silica-based slurry The effectiveness of the colloidal silica-based slurry on the reduction of the micro-scratch was evaluated by comparing it with the commonly used alumina-based slurry and the filtered alumina-based slurry. From the results of the colloidal silica-based slurry CMP experiments, the micro-scratch size was small when Rv ≤ 10 nm and the removal rate was as large as possible. Thus, the Al CMP was carried out at pH 4, H2 O2 1.0 vol.%, silica 5.0 wt.%. The PR CMP was conducted at pH 11, silica 0.5 wt.%. In previous studies, fine surfaces were obtained by the alumina slurry CMP when the Al CMP was performed at pH 2, H2 O2 5.0 vol.%, alumina 0.5 wt.% [15] and the PR CMP was at pH 11, alumina 0.5 wt.% [14]. ␥-Alumina of 0.05 m diameter was purchased from Buhler Co. and 1 m filter (CMP3) was provided by Millipore Co. The polishing times of Al and PR were 4 and 1 min, respectively. The Rv values of the Al and PR surfaces before the CMP were 39.4 and 7.85 nm, respectively. The results of the Al CMP and PR CMP experiments are shown in Table 1. The size (Rv) of the micro-scratch
Table 1 Summary of the CMP process results Thinfilm
Abrasive
Rv (nm)
RR (nm/min)
Al Al Al PR PR PR
Al2 O3 Filtered Al2 O3 Colloidal SiO2 Al2 O3 Filtered Al2 O3 Colloidal SiO2
18.15 9.85 6.08 13.31 5.96 15.27
21.34 2.57 137.45 1874.95 248.01 203.01
decreased, but the removal rate also decreased when the filtered alumina slurry instead of unfiltered slurry was used. The reason may be that the most of abrasive particles that were larger than 1 m were removed by the fine filter. Consequently, the large micro-scratches usually caused by the large particles occurred less and quick removal by the large particles could not occur. These phenomena are also well described by the CMP mechanism proposed by Luo and Dornfeld [16]. In the Al CMP, the micro-scratch size (Rv) and the removal rate were best when the colloidal silica-based slurry was used.Fig. 6 shows the Al film surfaces scanned by AFM (Cp autoprobe from PSI Co.) The scanned area was 50 m × 50 m. The Al film surface before CMP was relatively rough without any micro-scratches
Fig. 6. AFM images of Al films (a) before CMP and after CMP with (b) Al2 O3 , (c) filtered Al2 O3 , and (d) SiO2 abrasives.
Y. Ahn et al. / Wear 257 (2004) 785–789
(Fig. 6(a)). After the unfiltered alumina-based slurry CMP, the Al film had a fine surface, but micro-scratching occurred (Fig. 6(b)). The filtered alumina-based slurry CMP produced a finer surface than the unfiltered slurry, but it still had a few micro-scratches (Fig. 6(c)). The Al surface polished by the colloidal silica-based slurry had a very fine surface with few micro-scratches (Fig. 6(d)). The reason may be that, when the colloidal silica is smaller and has a lower hardness than alumina, the micro-scratches are reduced because the indented depth on the film surface decreases. The removal rate should be lessened in order to reduce the amount of micro-scratching when the Al CMP carried by the alumina-based slurry. Nonetheless, in the Al CMP by the colloidal silica-based slurry, if the slurry pH and the oxidizer concentration are chosen optimally, it may be able to obtain the fine surface without the micro-scratches while the removal rate is not small. In Table 1, the colloidal silica-based slurry showed the worse results than alumina-based slurry in the PR CMP. In order to obtain good performance in the organic polymer film CMP by the colloidal silica-based slurry, the more studies are needed. For example, a chemical solution that provides the polished surface with few micro-scratches and the improved removal rate must be researched.
4. Conclusions The following results were obtained from the Al and PR CMP experiments using the colloidal silica-based slurry. In the Al CMP, the micro-scratches were small and the removal rates were large when the slurry pH was 2–4 and the oxidizer H2 O2 concentration was 1–3 vol.%. The micro-scratch size and removal rate increased as the abrasive concentration in the slurry increased. In the PR CMP, the micro-scratch and removal rate generally increased as the slurry pH and abrasive concentration increased. The optimum CMP conditions from the colloidal silicabased slurry experiments and the optimum CMP were compared with the alumina-based slurry CMP. The Al CMP using the silica slurry showed a good result, but the PR CMP did not. The colloidal silica-based slurry produced a desirable fine Al surface with few micro-scratches which was comparable to that produced by a filtered alumina-based slurry.
789
Acknowledgements This work was supported by Korea Research Foundation Grant (KRF-2000-041-E00093). References [1] W. Jeffrey, L.D. David, W.L. Guthrie, J. Hopewell, F.B. Kaufman, W.J. Patrick, K.P. Rodbell, R.W. Pasco, A. Nenadic, US Patent 4,954,142, 1990. [2] W.J. Patrick, W.L. Guthrie, C.L. Standley, P.M. Schiable, Application of chemical mechanical polishing to the fabrication of VLSI circuit interconnections, J. Electrochem. Soc. 138 (6) (1991) 1778–1784. [3] J.J. Sniegowski, Chemical-mechanical polishing: enhancing the manufacturability of MEMS, Proc. SPIE—Micromachining Microfabrication Process Technol. II (1996) 104–115. [4] J.M. Steigerwald, S.P. Murarka, R.J. Gutmann, Chemical Mechanical Planarization of Microelectronic Materials, John Wiley and Sons, 1997, pp. 39 and 281. [5] L. Peters, Investigating the causing of CMP micro-scratches, Semicond. Int. 22 (6) (1999) 70. [6] G.B. Basim, J.J. Adler, U. Mahajan, R.K. Singh, B.M. Moudgil, Effect of particle size of chemical mechanical polishing slurries for enhanced polishing with minimal defects, J. Electrochem. Soc. 147 (9) (2000) 3523–3528. [7] C.G. Kallingal, D.J. Duquette, S.P. Murarka, An investigation of slurry chemistry used in chemical mechanical planarization of aluminum, J. Electrochem. Soc. 145 (6) (1998) 2074–2081. [8] L. Zhong, J. Yang, K. Holland, J. Grillaert, K. Devriend, N. Heylen, et al., A static model for scratches generated during aluminum chemical-mechanical polishing process: orbital technology, Jpn. J. Appl. Phys. 38 (1999) 1932–1938. [9] J.M. Steigerwald, S.P. Murarka, R.J. Gutmann, Chemical Mechanical Planarization of Microelectronic Materials, John Wiley and Sons, 1997, p. 49. [10] T. Hara, T. Tomisawa, T. Kurosu, T.K. Doy, Chemical mechanical polishing of polyarylether low dielectric constant layers by manganese oxide slurry, J. Electrochem. Soc. 146 (6) (1999) 2333– 2336. [11] S. Kondo, N. Sakuma, Y. Homma, Y. Goto, N. Ohashi, H. Yamaguchi, et al., Abrasive-free polishing for copper damascene interconnection, J. Electrochem. Soc. 147 (10) (2000) 3907–3913. [12] V.H. Bulsara, Y. Ahn, S. Chandrasekar, T.N. Farris, Mechanics of polishing, ASME J. Appl. Mech. 65 (1998) 410–416. [13] C.C. Yu, T.T. Doan, A.E. Laulusa, US Patent, 5,209,816, 1993. [14] H.-G. Kim, Y. Ahn, D.-K. Moon, J.-G. Park, slurry particles on chemical mechanical polishing of polyimide, Jpn. J. Appl. Phys. 39 (2000) 1085–1090. [15] W. Cho, Y. Ahn, C.-W. Baek, Y.-K. Kim, Effect of mechanical process parameters on chemical mechanical polishing of Al thin films, Microelectron. Eng. 65 (2003) 13–23. [16] J. Luo, D.A. Dornfeld, Effects of abrasive size distribution in chemical mechanical planarization: modeling and verification, IEEE Trans. Semicond. Manuf. 16 (3) (2003) 469–476.
