High-selectivity damascene chemical mechanical polishing
Shao-Yu Chiu
a, Ying-Lang Wang
b,c,*, Chuan-Pu Liu
d, Shih-Chieh Chang
c,
Gwo-Jen Hwang
b, Ming-Shiann Feng
a, Chia-Fu Chen
aaInstitute of Materials Science and Engineering, National Chiao-Tung University, Hsinchu 300, Taiwan, ROC bCollege of Science and Engineering, National University of Tainan, Taiwan
cDepartment of Applied Physics, National Chiayi University, Chiayi, Taiwan, ROC dDepartment of Material Science and Engineering, National Cheng kung University, Tainan, Taiwan
Available online 15 August 2005
Abstract
In this study, multi-step chemical mechanical polishing (CMP) with different copper removal rates and polishing pads is used to eliminate topography efficiently and to reduce micro-scratches on copper films. In colloidal-silica-based slurry, the polishing behaviors of copper, tantalum and silicon dioxide are found to relate to that kind of alkaline additives. The size of cations from alkaline additives influences the zeta potential of slurries, so as to vary the material removal rate. The addition of small-sized K+from KOH provides high removal selectivity of tantalum/copper and oxide/copper, so as to benefit the reduction of copper dishing.
D 2005 Elsevier B.V. All rights reserved.
Keywords: Chemical mechanical polishing; Micro-scratch; Tantalum; Copper
1. Introduction
In copper (Cu) damascene processes, chemical mechan-ical polishing (CMP) has been a promising method for achieving a globally planarized surface and defining interconnection areas [1 – 5]. CMP needs to produce a planarized Cu line with an intact surface because the quality of Cu surfaces directly contributed to the adhesion of dielectric films, the continuity of multilevel interconnection, and even stress reliability [6 – 11]. Chemical effect is particularly important in Cu-CMP because Cu is easily corroded and oxidized. Various abrasives have been developed to remove Cu oxides or passivation films, however, scratches are usually generated on Cu surfaces after CMP due to complicatedly tribochemical behaviors of Cu in different slurries and polishing conditions. Abrasives should be small or soft enough to avoid scratching while large or hard abrasives are needed to maintain a significantly
removal rate. Ideally, the Cu-CMP process should remove the excess Cu from barrier surfaces without losing Cu metals in trenches. A higher removal rate of Cu is required for the phase 1 Cu-CMP to efficiently eliminate the step height of the Cu films, whereas a lower removal rate of Cu is necessary for the phase 2 Cu-CMP to precisely control the end-point over the barrier metals. To guarantee the formation of the isolated interconnection lines, overpolish-ing is usually carried out to remove residual metals on the field surface outside of patterns. A major issue that remains is the severe dishing of Cu lines and erosion of dielectric films in high pattern density areas during the overpolishing step. To reduce the amount of Cu dishing, various slurry chemistries are investigated in this study. Colloidal silica is adopted as the abrasive in the second-step CMP, considering its high suspension stability in slurries with a wide pH range and its sphere shape with less damage on polished substrates. The polishing behaviors of Cu, tantalum (Ta), and silicon dioxide (SiO2) films in the colloidal-silica-based
slurries with hydrogen peroxide (H2O2) and alkaline
additives are discussed. Also, the formulation of slurries is optimized to improve the removal selectivity of different materials.
0040-6090/$ - see front matterD 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.tsf.2005.07.063
* Corresponding author. College of Science and Engineering, National University of Tainan, Taiwan. Tel.: +886 6 5051400x3031; fax: +886 6 5051273.
E-mail address: ylwang@tsmc.com (Y.-L. Wang).
Thin Solid Films 498 (2006) 60 – 63
2. Experimental details
All polishing experiments were performed with (100) oriented, 6-in.-diameter p-type Si wafers. The polishing sample was stacked with 900-nm-thick Cu and 50-nm-thick Ta, which was sputtering deposited on a 550-nm-thick thermal grown SiO2 wafer. For CMP analyses, the
thick-nesses of the Cu and Ta films were calculated by dividing the film resistivity with its measured sheet resistance. A Tencor M-gauge was used to measure the sheet resistance of the Cu and Ta films before and after CMP processes. The thickness of thermal SiO2film was measured by a Nanospec
210XP-thickness measurement system. Cu-CMP was performed with a Westech Model 372M polisher using a Politex Regular E. pad. Polishing parameters, such as down force, platen/ carrier rotary speed, back pressure and slurry flow rate were set to be 5 psi, 42/45 rpm, 2 psi and 150 ml/min, respectively. The polishing slurries under investigation were formu-lated with 10 wt.% diluted colloidal silica (< 50 nm in size), 10 wt.% H2O2(30 wt.% semiconductor grade) and adjusted
slurry pH value by adding various alkaline additives. Three types of alkaline additives, including potassium hydroxide (KOH), ammonium hydroxide (NH4OH) and
tetra-methyl-ammonium hydroxide (TMAH) were used as slurry pH adjusters. All chemical constituents mentioned above were added to the solution of de-ionized water with rigorous string.
In a colloidal dispersion of particles, there will generally be a net charge at the particle – liquid interface. The existence of the surface charge gives rise to the formation of the zeta potential. The zeta potential of the colloidal silica in the formulated slurries was measured by an electro-acoustic instrument, model DT1200. A laser diffraction particle size analyzer, UP-150 was used to measure the aggregation size of abrasives in slurries.
3. Results and discussion
The removal of Cu is contributed from both mechanical and corrosive (chemical) wears. In this study, the effects of slurry pH and added alkaline ions on the polishing behaviors of Cu, Ta and SiO2 are discussed. The
planarization performance of Cu-CMP is related to the properties of polishing pads. A soft pad, which has a smaller hardness and a larger compressibility, is usually used to prevent surface scratching, while a hard pad is usually used to reduce Cu dishing because its limited compressibility has a higher Cu removal rate on the feature tops. Fig. 1 shows the polishing behavior of Cu, Ta and SiO2 for various pH values. The slurry formulation
contains 10 vol.% H2O2, 10 wt.% colloidal silica, and
adding KOH to adjust slurry pH ranged from 2 to 10. In acidic aqueous medium, Cu tends to be corroded and then the removal rate of Cu is faster than those of Ta and SiO2
on which stable native oxide is present. The maximum
removal rate of Cu is above 130 nm/min at pH 2. As slurry pH increases into neutral regime, the removal rate of Cu drops as the dissolution rate of Cu is dramatically reduced by formation of Cu oxide passivation (Cu2O,
CuO). The solubility of Cu oxide becomes large as the slurry pH in the alkaline regime due to the formation of soluble Cu hydroxides or the decline of the hydrogen peroxide (H2O2) activity in the strongly alkaline solution.
Competing with the passivation behavior of Cu, the foregoing reactions result in few Cu removal rate increase (¨ 40 nm/min), as shown in Fig. 1.
In order to achieve global planarization, there should be high removal selectivity between metal, diffusion barrier and dielectric layer. So the polishing behavior of Ta and SiO2 should be noticed. For polishing Ta, it is found that
higher removal rates occur at both strongly acidic (pH 2) and alkaline (pH 10) slurries, as shown in Fig. 1. The Pourbaix diagram of the Ta – water system indicated the formation of tantalum pentoxide (Ta2O5) in a wide pH
range. The formation reaction of Ta2O5 on the Ta surface
could be expressed as follows [12]:
2Ta þ 5H2O ¼ Ta2O5 þ 10Hþ þ 10e: ð1Þ
In acidic solution, the oxygen reduction reaction may take place and be written as:
O2 þ 4Hþ þ 4e ¼ 2H2O: ð2Þ
The overall reaction can be rewritten as
2Ta þ 5=2O2 ¼ Ta2O5: ð3Þ
It has been reported that the presence of H2O2plays a
crucial role in determining the kinetics of the formation and dissolution of Ta2O5. The overall dissociation reaction of
H2O2is depicted as
2H2O2 ¼ 2H2Oþ O2: ð4Þ
H2O2is highly unstable in the alkaline solution. The rate
of the dissociation reaction of H2O2is enhanced in solutions
with high pH. The produced oxygen from the dissociation of
2 4 6 8 10 0 50 100 150 200 250 non-selectivity C u/Ta ~1 Inverse-selectivity Cu/Ta<1
10 wt.% colloidal silica with 10 vol.% H2O2
Cu Ta Oxide
Removal rate (nm/min)
pH value
Fig. 1. Removal rates of Cu, Ta, and oxide with various pH adjusted by KOH in the slurry with 10 wt.% colloidal silica slurry and 10 vol.% H2O2.
H2O2is suggested to facilitate noble Ta2O5forming soluble
hydroxotantalate and oxotantalate as shown in the following reactions[12]:
Ta5þ þ x=2H2Oþ x=4O2 þ xe ¼ TaðOHÞð5xÞþx ð5Þ
Ta5þ þ x=2H2Oþ ðx þ 2Þ=4O2 þ ðx þ 2Þe
¼ TaOðOHÞð3xÞþx : ð6Þ
In alkaline solutions, therefore, the removal rate of Ta is determined by the dissolution rate of Ta2O5 and increased
with increasing the pH value of the slurries.
For polishing Ta in solutions with low pH, the mechanical abrasion dominates the Ta removal rate due to the presence of stable Ta2O5 films in the acidic solutions.
Therefore, the larger secondary ionic particle size of abrasive at pH 2 leads to higher Ta removal rate because the iso-electric potential of colloidal-based slurries was around pH 2 in this study.
The removal behavior of oxide with slurry pH values is similar to that of Ta. As oxide CMP is carried out in acidic and neutral region, the removal of oxide is dominated by the mechanical abrasion, indicating that the larger abrasive at pH 2 induces higher oxide removal rate. In strongly alkaline slurries, the removal rate of oxide increases as the slurry pH increases because the hydrolysis reaction of oxide increases with increasing the slurry alkaline. In addition, adding more KOH into slurry would increase the ionic strength of slurry leading to a dramatic decrease in the thickness of diffusion layer. That means the repulsive electrostatic charges on both the abrasives and the substrate would be neutralized by the excess K+ cations, so the removal rates would be promoted.
Various kinds of alkaline additives, including KOH, NH4OH and TMAH, are investigated for their influence
on the removal behavior of Cu, Ta and SiO2 during
polishing. Fig. 2 shows the removal rates of Cu, Ta and SiO2 in the colloidal-silica-based slurries with various
alkaline additives (KOH, NH4OH, and TMAH) to control
pH value at 9.2.
It is through ammonium hydroxide that cupric ions would chelate to form soluble Cu(NH3)42+ and to increase
the Cu removal rate. As a result, polishing in the NH4
OH-containing slurry, the removal rate of Cu is increased to 225 nm/min, as shown inFig. 2. But the removal rate of Cu in the slurries with the other two alkaline additives (KOH and TMAH) is less than 30 nm. For Ta and SiO2 polishing in
the slurry with KOH, the maximum removal rates of Ta and SiO2are about 135 and 85 nm/min, respectively, but only
about 55 and 45 nm/min, respectively, in TMAH-added slurry at the same pH value. It implies that cations from the alkaline additives, i.e. K+, NH4+and N(CH3)4+, contribute to
the overall removal rate of Ta and SiO2. In general,
electrostatic double layer is within tens of micrometers. For smaller abrasives with a diameter of less than 50 nm, there is an electrostatic repulsive force between the negatively charged silica abrasives and the substrate. According to the theory of Stern layer, counter ions adsorption would decrease the surface potential. The negatively charged colloidal silica abrasives would be neutralized by adsorp-tion of caadsorp-tions from the alkaline additives. If the size of cations is large, there will be less cations adsorbed on the colloidal silica due to the effect of steric hindrance, so as to extend more the diffusion layer thickness. As shown inFig. 3, the absolute zeta potential (or the repulsive force) of colloidal silica in TMAH-added slurry is larger than that in KOH-added slurry because N(CH3)4+is larger than K+. The
high zeta potential of TMAH-added slurry leads to less impacting frequency of colloidal silica onto the substrate, and then the removal rate decreases.
For the NH4OH-based slurry, it is different from the other
alkaline solutions because NH3 would dissolve Cu oxide
passivation and form Cu(NH3)42+ to enhance the removal
rate of copper. The removal rate is about 225 nm/min for Cu and 80 nm/min for Ta at pH 9.2. With more NH4OH added
into the slurry, at pH 10, the removal rate would be enhanced to 1490 nm/min for Cu and 100 nm/min for Ta. So 0 500 1000 1500 2000 2500 Cu Ta TMAH NH4OH KOH Oxide
Removal Rate (A/min)
Fig. 2. Polishing rate of Cu in colloidal-silica-based slurry with H2O2using
various alkaline additives to adjust pH value at 9.2.
20 40 60 80 100 -60 -40 -20 0 TMAH NH4OH Zeta potential (mV)
Mean particle size zeta potential
Mean particle size (nm)
KOH
Fig. 3. Mean particle size and zeta potential of colloidal-silica-based slurry with H2O2using various alkaline additives to adjust pH value at 9.2.
S.-Y. Chiu et al. / Thin Solid Films 498 (2006) 60 – 63 62
the removal selectivity of Cu/Ta and Cu/ SiO2gets higher
(14.3 and 19, respectively) and that would be suitable for the 1st step of Cu removal but not for the 2nd step of Ta removal.
In order to reduce metal dishing and oxide erosion, it is better to control fast removal rate of Ta and not to remove Cu and SiO2, respectively. The KOH-added slurry has
higher removal selectivity of Ta/Cu and SiO2/Cu.
Further-more, the Ta removal rate is relatively high (¨ 140 nm/min) in the KOH-added slurry. So KOH is found to be a suitable pH adjuster for the colloidal-silica-based slurry.
4. Conclusions
In this study, the removal mechanism of Cu-CMP on the characteristics of colloidal-silica-based slurry is investi-gated. It is found that the polishing behavior of Cu, Ta and oxide is deeply dependant on the slurry pH and the surface-charge-related interactions between the slurry and the polishing substrate. The attraction and expulsive force affect the impacting frequency between colloidal silica abrasives and the wafer surface. Large-size cations lead to a less impacting frequency between the abrasives and the sub-strate, so as to get lower removal rate. Combining the results mentioned above, high or low removal selectivity of Ta/Cu and SiO2/Cu during the two-step Cu-CMP process can be
achieved by modified colloidal-silica-based slurries with proper alkaline additives and pH values.
Acknowledgments
The authors would like to thank the National Science Council of the Republic of China, Taiwan, for partially supporting this research under Contract no. NSC92-2216-E006-046. Technical assistance from the National Nano Device Laboratories of NSC is also appreciated.
References
[1] J. Hernandez, P. Wrschka, G.S. Oehrlein, J. Electrochem. Soc. 148 (2001) G389.
[2] S. Aksu, F.M. Doyle, J. Electrochem. Soc. 149 (2002) G352. [3] S. Tamilmani, W. Huang, S. Raghavan, R. Small, J. Electrochem. Soc.
149 (2002) G638.
[4] D. Ernur, S. Kondo, K. Maex, Jpn. J. Appl. Phys. 41 (2002) 7338. [5] T.C. Hu, S.Y. Chiu, B.T. Dai, M.S. Tsai, I.C. Tung, M.S. Feng, Mater.
Chem. Phys. 61 (1999) 169.
[6] S. Varadarajan, D. Kalakkad, T. Cacouris, Semicond. Int. 25 (2002) 6. [7] A. Beverina, H. Bernard, J. Palleau, J. Torres, F. Tardif, Electrochem.
Solid-State Lett. 3 (2000) 156.
[8] Y. Homma, S. Kondo, N. Sakuma, K. Hinode, J. Noguchi, N. Ohashi, H. Yamaguchi, N. Owada, J. Electrochem. Soc. 147 (2000) 1193. [9] S. Kondo, N. Sakuma, Y. Homma, N. Ohashi, Jpn. J. Appl. Phys. 39
(2000) 6216.
[10] D. Zeidler, Z. Stavreva, M. Plotner, K. Drescher, Microelectron. Eng. 37/38 (1997) 237.
[11] D. Ernur, S. Kondo, D. Shamiryan, K. Maex, Microelectron. Eng. 64 (2002) 117.
[12] S.C. Kuiry, S. Seal, W. Fei, J. Ramsdell, V.H. Desai, Y. Li, S.V. Babu, B. Wood, J. Electrochem. Soc. 150 (2003) C36.