Highly reliable chemical–mechanical polishing process for organic low-k
methylsilsesquioxane
Po-Tsun Liu, Ting-Chang Chang, Ming-Chih Huang, M. S. Tsai, and S. M. Sze
Citation: Journal of Vacuum Science & Technology B 19, 1212 (2001); doi: 10.1116/1.1385684 View online: http://dx.doi.org/10.1116/1.1385684
View Table of Contents: http://scitation.aip.org/content/avs/journal/jvstb/19/4?ver=pdfcov Published by the AVS: Science & Technology of Materials, Interfaces, and Processing
Articles you may be interested in
Remote H2/N2 plasma processes for simultaneous preparation of low-k interlayer dielectric and interconnect copper surfaces
J. Vac. Sci. Technol. B 30, 031212 (2012); 10.1116/1.4705732
Electrical properties of organic light-emitting diodes by indium tin oxide chemical-mechanical polishing process J. Vac. Sci. Technol. A 25, 999 (2007); 10.1116/1.2746051
Method to improve chemical-mechanical-planarization polishing rate of low-k methyl-silsesquiazane for ultralarge scale integrated interconnect application
J. Vac. Sci. Technol. B 22, 1196 (2004); 10.1116/1.1755218
Effective repair to ultra-low-k dielectric material (k2.0) by hexamethyldisilazane treatment J. Vac. Sci. Technol. B 20, 1334 (2002); 10.1116/1.1488645
Effects of H 2 plasma treatment on low dielectric constant methylsilsesquioxane J. Vac. Sci. Technol. B 17, 2325 (1999); 10.1116/1.590912
for organic low-k
methylsilsesquioxane
Po-Tsun Liua)and Ting-Chang Changb)National Nano Device Laboratory, 1001-1 Ta-Hsueh Road, HsinChu 300, Taiwan, Republic of China Ming-Chih Huang
Institute of Electronics, National Chiao Tung University, Taiwan, Republic of China M. S. Tsai and S. M. Szec)
National Nano Device Laboratory, 1001-1 Ta-Hsueh Road, HsinChu 300, Taiwan, Republic of China 共Received 19 May 2000; accepted 21 May 2001兲
In this work, chemical–mechanical polishing共CMP兲 of the organic polymer, methylsilsesquioxane
共MSQ兲, has been investigated. For conventional silicate-based slurry, the CMP removal rate of MSQ
is low and many scratches are formed at the surface. Moreover, the dielectric properties of a post-CMP MSQ film are degraded in comparison to the as-cured MSQ. We have proposed a reliable process for the CMP of MSQ which includes a slurry of additive and a post-CMP NH3 plasma treatment. Experimental results show that the modified slurry provides a high polishing rate and uniform surface topography. In addition, the NH3 plasma process can form a thin nitrogen-containing layer on the post-CMP MSQ surface, which enhances the resistance to moisture absorption and copper diffusion. © 2001 American Vacuum Society. 关DOI: 10.1116/1.1385684兴
I. INTRODUCTION
As devices continuously scale down, interconnect delay becomes the performance barrier for high-speed conduction. Insulating dielectrics with low permittivity 共low-k兲1–3 have been popularly applied to multilevel interconnect architec-ture reducing signal propagation time delay, cross talk, and power consumption issues. In general, organic polymers are known to exhibit lower dielectric constants than inorganic oxides and nitrides, and may be considered as candidates for intermetal dielectric 共IMD兲 in the ultralarge scale integrated circuit era.4,5However, polymer dielectrics still have several process integration issues that must be addressed prior to implementation. The biggest concern for low-k polymer ap-plications is the ‘‘poisoned-via’’ issue.6An etchback step is commonly performed to solve this issue. By removing the low-k films from where the vias will subsequently be formed, the possibility of moisture outgassing from the low-k dielectrics is effectively prevented.7 Therefore, surface pla-narization after etchback processing is a key technology dur-ing the manufacturdur-ing of multilevel interconnects. The chemical–mechanical planarization 共CMP兲 process is satis-factory to achieve global topography planarization. Due to the hydrophobic surface of organic low-k polymers, CMP is a difficult process. Many deep scratches are formed on the organic surface when these dielectrics are polished mechanically.8–11 These defects are harmful for the perfor-mance of the device, so care must be taken not to damage the polymer IMD during the polishing of the low-k film.
In this work, we will present a reliable CMP process for
organic low-k methylsilsesquioxane 共MSQ兲. Methylsilses-quioxane having the general formula (CH3SiO1.5)2n, n
⫽2, 3, etc. belongs to the polymeric family of silicones.12 The silicones are polymers containing an inorganic backbone of alternating silicon and oxygen atoms with organic groups attached to silicon. MSQ, which exhibits a relatively low dielectric constant (k⫽2.6– 2.8) as compared to SiO2(k
⫽4.0), is intrinsically hydrophobic, and has reasonable
me-chanical hardness, and possesses exceptional thermal and di-mensional stability.13–15For these reasons, MSQ represents an excellent candidate for applications on the multilevel in-terconnect architecture as IMD.
The polishing of MSQ film with conventional silica-based slurry is discussed in this study. Additionally, tetramethy-lammonium hydroxide共TMAH兲 was added to a conventional silica-based slurry and its effect on the polish rate was inves-tigated. This study has employed material analyses and elec-trical measurements to characterize the post-CMP MSQ. Fi-nally, a NH3 plasma technique was implemented and characterized as a post-CMP treatment for the processing of low-k MSQ.
II. EXPERIMENT
The precursor for the films evaluated in this study was X418, a commercially available solution of methyl silsesqui-oxane in ethanol, n-butanol, propylene, glycol methyl ether acetate manufactured by Allied Signal Corporation. MSQ films were prepared by spin-coating X418 solution onto 6 in.
p-type 共25 ⍀ cm兲 single crystal silicon wafers with 共100兲
orientation. Spin speed was 3000 rpm and the spinning time was 20 s, which gave a MSQ thickness about 500 nm. Then, the wafers were baked sequentially on a hot plate at 180 °C for 2 min and 250 °C for 1 min. The resulting wafers were cured in furnace at 400 °C for 30 min under nitrogen
ambi-a兲Electronic mail: ptliu@ndl.gov.tw
b兲Also at: Department of Physics, National Sun Yat-Sen University, Taiwan,
R.O.C.
c兲Also at: Institute of Electronics, National Chiao Tung University, Taiwan,
ent. It was followed that as-cured MSQ films were processed through CMP. The CMP experiment was carrier out on an IPEC/Westech 372M CMP processor consisting of a Rodel IC 1400 pad on the primary polishing platen, a Rodel Politex Regular™ embossed pad on the final buffing platen, a carrier to hold wafers against the pad, and a Rodel R200-T3 carrier film to provide buffer between the carrier and the wafer. The wafer was mounted on a template assembly for a single 6 in. wafer during the polishing experiment. The most commonly used slurry for SiO2 polishing is silica with potassium hy-droxide共KOH兲 aqueous solution, 30 wt % solid content, pH 10.4, mean particle size 110 nm, named CABOT SS-25™ slurry. Since low-k MSQ is one group of siloxane-based SOG films, the slurry used in this work was SS-25 slurry diluted 1:1 with deionized water. In parallel, work was de-veloped to increase the CMP removal rate and uniformity of MSQ in silica-based slurry by adding 0.1–0.2 M TMAH
aqueous solution to commercial SS-25 slurry. The resultant solution pH is in the range of 11–12. The optimum polishing parameters, such as down force, back pressure, platen and carrier rotation speeds, and slurry flow rate, were set to be 3 psi, 2 psi, 50 rpm, 60 rpm, 150 ml/min, respectively. The thickness and refractive index of all the blanket films before and after CMP polishing were measured using an n and k 1200 Analyzer, with 12 points measured.
Subsequently, the polished wafers were transferred to plasma enhanced chemical vapor deposition chamber for the NH3plasma posttreatment. The NH3plasma was operated at a pressure of 300 mTorr and with a NH3gas flow rate of 700 sccm. A rf power of 200 W, which established the NH3 plasma, was applied to the upper electrode and the wafers were placed on the bottom by a grounded electrode, which can be rotated for improving uniformity, at a substrate tem-perature of 300 °C. The structural properties of the MSQ
FIG. 1. 共a兲 Variation in the removed thickness of CMP MSQ vs polish time. A large variation in the film thickness of MSQ is observed. 共b兲 Top view 共left兲 and surface line scanning profile 共right兲 by AFM measurement for the surface of polished MSQ with silicate-based SS-25 slurry only. A poor surface topography is observed. The Ra within an image area of 100m2is 1.32 nm.
1213 Liuet al.: Highly reliable chemical-mechanical polishing 1213
films were studied using Fourier transform infrared absorp-tion spectra 共FTIR兲. The surface morphologies of the pol-ished films were investigated by atomic force microscopy
共AFM兲 with a scan area 100m2. Material analysis such as thermal desorption system spectrometer 共TDS兲 was carried out to monitor the desorbed moisture from the post-CMP MSQ films during a high temperature process. Electrical characterizations of post-CMP MSQ films were performed on the metal–insulation–semiconductor capacitors with me-tallic aluminum deposition as top electrode. Leakage current–voltage and capacitance–voltage characteristics were also used to analyze the leakage current behavior and measure dielectric constant of post-CMP MSQ film, respec-tively. For all electrical measurements, 21 dies were mea-sured on per wafer and 50 pieces of wafers in total.
III. RESULTS AND DISCUSSION
For the initial experiments, the CMP process was imple-mented with commercial silica-based slurry SS-25™ since the MSQ is one of the silicon-based low-k materials. Figure 1共a兲 illustrates the variation of removed thickness versus pol-ish time. The results show that the polpol-ishing rate of CMP organic MSQ is rather low共about 100 Å/min兲. In addition, a large variation in the film thickness of MSQ is observed. AFM images were used for further observation of the result-ant topography after CMP, as shown in Fig. 1共b兲. The mean surface roughness 共Ra兲 within an image area of 100m2 is even up to 1.32 nm as compared to 0.154 nm of pre-CMP surface. This difficulty in polishing is also consistent with Forester et al.11 which presents that the organic content in the dielectric films will inhibit the hydration reaction during
FIG. 2. 共a兲 Removed thickness of CMP MSQ with and without additive TMAH for the same polishing parameters of CMP process. The removal rate is 1200 and 100 Å/min, respectively.共b兲 Top view 共left兲 and surface line scanning profile 共right兲 by AFM measurement for the surface of polished MSQ with TMAH-added slurry. The surface topography of polished MSQ film is smooth. The Ra within an image area of 100m2is 0.235 nm.
CMP. Thus, the polishing of polymers can be regarded as a chemical reaction limited process rather than mechanically limited. In order to increase the polishing rate, the chemical surfactant TMAH which can improve the MSQ surface wet-tability is mixed with SS-25™ as the modified slurry in this study. Figure 2共a兲 shows the removed thickness of CMP MSQ with and without additive TMAH for the same polish-ing parameters. A higher MSQ removal rate 共about 1200 Å/min兲 is observed for the SS-25™ slurry mixing with addi-tive TMAH. The surface line scanning profile of polished MSQ with TMAH-added slurry is shown in Fig. 2共b兲. Com-pared to the polished surface with SS-25™ slurry only, the surface roughness of post-CMP MSQ with TMAH is signifi-cantly decreased to 0.235 nm. This implies a more uniform
dissolution and removal is occurring on the surface of or-ganic MSQ. In this work the surfactant TMAH which has both hydrophilic and hydrophobic groups in its molecular components, can reduce the surface energy between abrasive and the hydrophobic MSQ. In addition, the ammonium hy-droxide ion pairs will absorb on the organic MSQ surface resulting in the increased pH locally共from 10.4 to 11兲 on the surface than KOH ions do in the SS-25 slurry. This will facilitate the hydrolysis reactions of MSQ in a more basic environment. Both the improvement in MSQ surface wetta-bility and the increase in dissolution constant of hydroxide ions are the necessary factors to improve the CMP MSQ. The hydration reaction scheme has been proposed in a pre-vious article.16 Therefore, in aqueous solution the modified slurry (SS-25™⫹TMAH) is capable of dissolving the or-ganic content uniformly and initiating hydration reactions to breaking the Si–O bonds in the MSQ films. Meanwhile, the mechanical force during the CMP process provides another energy to enhance moving the slurry particles away, which is associated with the breakdown of MSQ backbone. With an enhanced hydration reaction and the assistance of the pro-ceeding mechanical event, a high removal rate of CMP MSQ can be obtained as compared to SS-25™ slurry only.
Furthermore, we investigate the electrical characteristics of the post-CMP MSQ film to evaluate the influence of the CMP process on low-k properties. Figures 3共a兲 and 3共b兲 show the leakage current and dielectric constant of MSQ after the CMP process with and without TMAH-added slurry. The leakage current densities of both cases of post-CMP MSQ are increased as much as one order of magnitude higher than that of pre-CMP MSQ. The dielectric constant of post-CMP MSQ is also increased from an as-cured value of 2.6–2.86. In addition, we observe that both the leakage cur-rent and dielectric constant of post-CMP MSQ with TMAH-added slurry are slightly increased when compared to the case of CMP MSQ without the TMAH additive. FTIR spec-tra show that the peak intensity of the Si–C bond and the C–H bond of post-CMP MSQ is reduced in comparison with that of pre-CMP MSQ, as shown in Fig. 4. The destruction of organic functional bonds from kinematic mechanical abra-sion and chemical reaction leads to dielectric deterioration and inevitably results in the increased leakage current and dielectric constant. This is especially true for the enhanced hydrolysis reaction of MSQ with the addition of TMAH which is responsible for even more degradations in electrical characteristics. This delitescent effect will offset partial ad-vantage of using additive TMAH.
To alleviate the electrical degradation, an NH3 plasma treatment was applied upon the post-CMP MSQ. Figures 5共a兲 and 5共b兲 show the leakage current density and dielectric constant of post-CMP MSQ treated with NH3plasma treat-ment for 3–9 min. These results show that both the leakage current and dielectric constant of post-CMP MSQ decrease and approach the pre-CMP values, with the increase in NH3 plasma treatment within 9 min. In this work, a NH3-plasma treatment within 10 min is an appropriate condition, or more nitridation would raise the dielectric constant of the MSQ
FIG. 3. Dielectric properties of MSQ polished with and without additive TMAH. 共a兲 Leakage current density of post-CMP MSQ as a function of electric field. 共b兲 Dielectric constant of post-CMP MSQ films. Both the leakage current and dielectric constant of post-CMP MSQ are higher than that of pre-CMP MSQ.
1215 Liuet al.: Highly reliable chemical-mechanical polishing 1215
film gradually. These significant improvements in electrical characteristics of post-CMP MSQ can be interpreted by ma-terial analyses. X-ray photoelectron spectroscopy 共XPS兲 analysis confirms the modification of the MSQ surface. Fig-ures 6共a兲 and 6共b兲 show XPS diagrams of the post-CMP MSQ film before and after NH3plasma treatment. Compared to Fig. 6共a兲, it is found that a significant signal of nitrogen appears at about 400 eV in Fig. 6共b兲. This means that the nitrogen atoms are shallowly doped in the MSQ. Jeong et al. demonstrated that the SiON phase is formed in the MSQ surface after NH3 treatment.17 In order to realize the SiON barrier against moisture uptake, TDS analysis was performed for further investigation. Figure 7 shows the temperature de-pendence of moisture desorption from post-CMP MSQ film with and without NH3plasma treatment. The moisture con-tent in NH3-plasma treated MSQ is lower than that of un-treated post-CMP MSQ. The inert passivation layer formed by NH3-plasma treatment effectively prevents moisture ab-sorption in post-CMP MSQ. The leakage current and dielec-tric constant of post-CMP MSQ are thereby decreased due to the reduction in polar moisture uptake.18,19 Additionally, NH3-plasma treatment can block copper penetration in
post-CMP MSQ. Secondary ion mass spectroscopy共SIMS兲 analy-sis was carried out to observe the distribution of copper ele-ment after removing the Cu electrode from the thermally stressed Cu/MSQ/Si capacitors. Figure 8 shows SIMS depth profile of copper in post-CMP MSQ with and without NH3-plasma treatment after being subjected to a thermal stress of 500 °C for 30 min under nitrogen ambient. The SIMS profile shows copper penetrates and piles up on the interface between the untreated MSQ film and the Si sub-strate under thermal stressing. This is a serious problem since the existence of copper in dielectrics leads to reliability
FIG. 4.共a兲 FTIR spectra of low-k MSQ film before and after CMP process. 共b兲 An enlargement of the 2800–3200 cm⫺1region for共a兲. The destruction of function groups共Si–C and C–H bonds兲 appear after CMP MSQ.
FIG. 5. 共a兲 Leakage current density of post-CMP MSQ with NH3-plasma treatment as a function of electric field.共b兲 Dielectric constant of post-CMP MSQ with NH3-plasma treatment as a function of treatment time. Both the leakage current and dielectric constant decrease and approach the pre-CMP values. An NH3-plasma treatment within 10 min is an appropriate condition, or more nitridation will raise the dielectric constant of MSQ film gradually.
issues.20,21By contrast, the profile of copper has a relatively shallow distribution which rapidly tails off in the NH3-plasma treated MSQ film due to the presence of nitro-gen. Therefore, the NH3-plasma treatment also effectively enhances the resistance of the post-CMP MSQ to copper penetration.
IV. CONCLUSIONS
The existence of the methyl group in a dielectric film will largely reduce the CMP removal rate and make CMP diffi-cult to achieve a uniform polish across the wafer when using conventional oxide CMP slurry. This work has reported an efficient CMP process for organic low-k MSQ as an
inter-metal dielectric material. The commercial SS-25™ silica-based slurry combined with the additive TMAH can acceler-ate the polish racceler-ate of organic MSQ film. Since the addition of TMAH is quite capable in converting the hydrophobic MSQ surface into a more hydrophilic condition, the removal rate of low-k MSQ is promoted from approximately 100–to 1200 Å/min. In addition, the application of a NH3-plasma
post-FIG. 6. 共a兲 Full XPS profile of as-deposited MSQ surface. 共b兲 Full XPS profile of NH3-plasma treated MSQ surface, showing the presence of ele-ment N.
FIG. 7. Temperature dependence of moisture desorption from post-CMP MSQ films with and without NH3-plasma treatment. The moisture content in NH3-plasma treated MSQ is lower than that of untreated post-CMP MSQ.
FIG. 8. SIMS depth profiles of copper in post-CMP MSQ film with and without NH3-plasma treatment after being subjected to a thermal stress of 500 °C for 30 min under nitrogen ambient. A relatively shallow copper distribution is observed in NH3-plasma treated MSQ in comparison to un-treated MSQ.
1217 Liuet al.: Highly reliable chemical-mechanical polishing 1217
CMP treatment will restore the dielectric degradation due to the mechanical abrasion and chemical etching reaction dur-ing the CMP process. XPS spectrum has shown the presence of nitrogen on the surface of an NH3-plasma treated post-CMP MSQ. The presence of nitrogen effectively prevents the post-CMP MSQ from moisture uptake and copper diffu-sion which have been confirmed by TDS and SIMS analysis, respectively.
ACKNOWLEDGMENTS
This work was performed at the National Nano Device Laboratory and was supported by Allied Signal Taiwan, Inc. and the National Science Council of the Republic of China under Contract No. NSC 90-2721-2317-200.
1International Technology Roadmap for Semiconductors共Semiconductor Industry Association, San Jose, CA, 1997兲.
2T. Nakano and K. Tokunaga, J. Electrochem. Soc. 142, 1303共1995兲. 3
P. T. Liu, T. C. Chang, Y. L. Yang, Y. F. Cheng, J. K. Lee, F. Y. Shih, E. Tsai, G. Chen, and S. M. Sze, J. Electrochem. Soc. 147, 1186共2000兲. 4A. K. Stamper, V. McGahay, and J. P. Hummel, Proceedings of the 3rd
International Dielectrics for ULSI Multilevel Interconnection Conference, DUMIC, Santa Clara, CA, 10–11 February 1997, p. 13.
5
M. Morgen, J. H. Zhao, M. Hay, T. Cho, and P. S. Ho, Mater. Res. Soc. Symp. Proc. 565, 69共1999兲.
6J. Wataerloos, H. Meynen, B. Coenegrachts, and L. Van den hove, No. 97-ISMIC-222D, Proceedings of the 3rd International Dielectrics for ULSI Multilevel Interconnection Conference, Santa Clara, CA, 18–20 June 1996, p. 52.
7L. Forester, S. B. Kim, D. K. Choi, J. H. Ko, J. C. Kim, M. Ross, W. R.
Livesay, and J. J. Yang, Proceedings of the 2nd International Dielectrics for ULSI Multilevel Interconnection Conference, Santa Clara, CA, 18–20 June 1996, p. 119.
8Y. Homma, T. Furusawa, H. Moroshima, and H. Sato, Solid-State Elec-tron. 41, 1005共1997兲.
9T. Hara, T. Tomisawa, T. Kurosu, and T. K. Doy, J. Electrochem. Soc. 146, 2333共1999兲.
10C. L. Borst, R. J. Gutmann, and W. N. Gill, Proceedings of the 16th International VLSI Multilevel Interconnection Conference, Santa Clara, CA, 7–9 September 1999, p. 207.
11L. Forester, D. K. Choi, R. Hosseini, J. Lee, B. Tredinnic, K. Holland, and T. Cale, Proceedings of the 12th International VLSI Multilevel Inter-connection Conference, Tampa, FL, 27–29 June 1995, p. 482. 12N. H. Hendricks, Mater. Res. Soc. Symp. Proc. 443, 3共1997兲. 13C. T. Chua, G. Sarkar, and X. Hu, J. Electrochem. Soc. 145, 4007共1998兲. 14P. T. Liu, T. C. Chang, Y. S. Mor, and S. M. Sze, Jpn. J. Appl. Phys. 38,
3482共1999兲. 15
A. T. Kohl, R. Mimna, R. Shick, L. Rhodes, Z. L. Wang, and P. A. Kohl, Electrochem. Solid-State Lett. 2, 77共1999兲.
16Y. L. Wang, C. Liu, S. T. Chang, M. S. Tsai, M. S. Feng, and W. T. Tseng, Thin Solid Films 308–309, 550共1997兲.
17H. D. Jeong, H. S. Park, H. J. Shin, B. J. Kim, H. K. Kang, and M. Y. Lee, Proceedings of the 1999 International Interconnect Technology Con-ference, San Francisco, CA, 24–26 May 1999, p. 190.
18A. L. S. Loke, J. T. Wetzel, J. J. Stankus, M. S. Angyal, B. K. Mowry, and S. S. Wong, IEEE Electron Device Lett. 19, 177共1998兲.
19R. A. Swalin, Thermodynamics of Solids共Wiley, New York, 1972兲, p. 302.
20
G. Raghavan, C. Chiang, P. B. Anders, S. M. Tzeng, R. Villasol, G. Bai, M. Bohr, and D. B. Fraser, Thin Solid Films 168, 262共1995兲. 21A. L. S. Loke, J. T. Wetzel, P. H. Townsend, T. Tanabe, R. N. Vrtis, M.
P. Zussman, D. Kumar, C. Ryu, and S. S. Wong, IEEE Electron Device Lett. 46, 2178共1999兲.
![TraditionalMLCalgorithmsmainlytacklethebatchMLCproblem,wheretheinputdataarepresentedinabatch[24,28].Nevertheless,inmanyMLCapplicationssuchase-mailcategorization[22],multi-labelexamplesarriveasastream.Onlineanalysisistherefore dimensionreducermotivatedbyma](data:image/gif;base64,R0lGODlhAQABAIAAAP///wAAACH5BAEAAAAALAAAAAABAAEAAAICRAEAOw==)