Post cleaning of chemical mechanical polishing process

ELSEVIER Applied Surface Science 92 (1996) 176-179

• ~..~: :i~.~:~'~:::~:,:,,-..~.. :~.:...~ ~ - : ~ ! ~ ; ~ : ~ . ~ , ~ . . ' : . ~ ~:~::::~.:.:... ..~.:.::.:~..:~:~

surface sc i enc e

Post cleaning of chemical mechanical polishing process

C h i - W e n L i u

a,

a Department o f Electronics Engineering & Institute o f Electronics, National Chiao Tung University, 1001 Ta Hsueh Road, Hsinchu 30050, Taiwan, ROC

b NationalNano Device Laboratory, 1001-1 Ta Hsueh Road, Hsinchu 30050, Taiwan, ROC

Received 12 December 1994; accepted for publication 29 May 1995

Abstract

We describe a study on the effect of the electrostatic nature in silica particles on the post CMP cleaning behavior. A fall-off for the zeta potential of silica particles is observed as the pH of dip solutions is increased. In this study, we also observed that particle counts on the SiO 2 and the Si3N 4 dielectric films had a similar dependence on the pH. Furthermore, we confirmed that surface hardness of the wafer is an important factor for particles physically embedded in different dielectric materials during and after the CMP process. The nanoscale surface hardness of dielectric films was measured by the nanoindentation technique. Experimental results showed that particles had difficulty attaching to a harder surface of the dielectric film.

1. Introduction

Since integrated circuit devices continue to be scaled down, global planarization is demanded as chip designs include multi-level interconnections with decreasing photolithographic depth of field. Tradi- tional planarization methods such as SOG and etch back, planarize over a distance in a few microns range. Currently, chemical-mechanical polishing (CMP) is the only method available that planarizes effectively at local and global range. Therefore, CMP has become an important technology for inter-level dielectric and shallow trench isolation planarization [1,2].

* Corresponding author. Tel.: + 886-35-726100 Ext. 7719; fax: + 886-35-713403.

The polishing process involves intimate contact between the wafer surface and the pad material in the presence of slurry. The colloidal debris from slurry will be left on the wafer surface after polish- ing. These residual silica will contaminate the subse- quent processing steps and cause functional defects and lower yield in the finished integrated circuit. Therefore, an effective wafer surface cleaning is important for transferring CMP process from R & D labs to manufacture sites directly.

The particles attached on the wafer surface come from both chemical and physical procedures. In the CMP process, silica particles with sufficient mass and energy can penetrate into the surface of the polished film, resulting in abrasion. As the carrier and wafer subsystem lift up and leave the pad, those silica particles which have penetrated into the pol- ished film but not removed in-situ during CMP 0169-4332/96/$15.00 © 1996 Elsevier Science B.V. All fights reserved

C.-W. Liu et al. /Applied Surface Science 92 (1996) 176-179 177

process will be embedded in the polished film. Since the hardness is determined by the depth of penetra- tion of an indenter forced (in our case, silica parti- cles) into a surface [3], the amounts of the silica particles embedded in polished films would vary with surface hardness. On the other hand, since this slurry is a suspension of fused silica dispersed in aqueous potassium hydroxide, the electrostatic forces, i.e. particle zeta potential, will control the particle deposition [4]. Kern [5] reported that colloidal depo- sition on a hydrophilic wafer surface exhibits a dependence on the pH similar to the zeta potential. Thus, silica particles adhered to the polished film would depend on the pH of the dip solution. How- ever, there was still a lack of experimental data to verify this point for the slurry used in the CMP process. In this article we reported on a comprehen- sive characterization of particle removal via in-situ dipping process which was performed with solutions of various pH values. We also demonstrated that hardness of the polished film was an essential factor in the post CMP cleaning behavior.

After the abrasive process, the wafer and carrier were submerged in a tub of deionized water immedi- ately, and then the wafer was transported by the carrier to unloaded cassette.

Cab-O-Sperse SC-1 slurry (a common type, avail- able from Rippey Corporation) was used for polish- ing. SC-1 is a suspension of fused silica dispersed in aqueous potassium hydroxide. The number of parti- cles ( > 0.2 ixm) on a wafer were counted with a Surface Scan (TENCOR-4500).

2.3. Zeta potential measurement

In a colloidal dispersion of particles there will generally be a net charge at the particle-liquid inter- face. The existence of this surface charge gives rise to the formation of the zeta potential [5,6]. The values of zeta potentials for silica particles were calculated by the Smoluchowski equation [6] using the mobility measurements (Zeta-meter system 3.0, Zeta-Meter, Inc.).

2.4. Nanoindentation measurement

2. Experimental procedure

2.1. Sample preparation

All test samples in this study were prepared on p-type (100), 150 mm silicon wafers. Thermally grown silicon dioxide films were processed by wet oxidation (ASM/LB45 furnace system), in which the silicon was exposed to the ambient H 2, 02 at 980°C. Si3N 4 film was deposited by the interactions of NH 3, N 2, N20 and SiH2C12 at 780°C (Vertex, DJ-802V, Kokusai Electric).

An indentation system (Nanotest 500, Micro Ma- terials, Inc.) with a Berkovich indenter was used to measure the nanoscale mechanical properties of di- electric films. All tests were performed at a nominal constant loading rate of 1.18 m N / s of the maximum set displacement (300 nm). For each sample, l0 separate indents, spaced 15 Ixm from each other, were made on the surface. At each load, the load versus displacement curve was recorded, from which hardness can be calculated using the standard for- mula [7].

2.2. CMP process 3. Results and discussions

A Westech Model 372M CMP System, consisting of a IC 1000/Suba IV (made of polyurethane im- pregnated polyester) pad affixed to a circular polish- ing platen and a carrier to hold wafers against the pad, was used for polishing. During the polishing operation, both platen and carrier were rotated at 20 and 42 rpm, respectively. We pressed the wafer in the carrier against the platen with a force of 7 psi, and the back pressure against the wafer was 2 psi.

As it is well known, the zeta potential can be altered by adjusting the pH. In our experiment, the pH of the slurry was adjusted with HNO 3 to various values, which were measured (with a precision of +_0.01) by a microprocessor pH meter (SP 2200, Suntex). Fig. 1 shows the change of the zeta poten- tial with the pH in SC-1 slurry. Also included is the zeta potential o n S i O 2 and SiaN 4 film surface by streaming potential measurements proposed by

178 C.-W. Liu et al./Applied Surface Science 92 (1996) 176-179

Bousse and coworkers [8,9]. The result shows that raising the pH will decrease the value of the zeta potential of the silica particles. The same tendency exhibits on the surface of the dielectric films [8,9]. The fact that both silica particles and film surfaces in liquids become electrically charged implies that elec- trostatic forces can affect the deposition process.

In the post CMP cleaning test, adjustment of the dip solution pH was carded out by adding NH4OH and HCI to deionized water. Particle counts on the SiO 2 and the Si3N 4 films immersed in various pH solutions for 30 min are shown in Fig. 2. A decrease in particle counts is observed as the pH value is increased except for the case of pH less than 7. The pH dependence of particle counts is very similar to the situation in the zeta potential of silica particles. Because adsorbed monomeric silica or colloid silica on the wafer may gel if it is exposed to a solution of pH 7 or less [10], the silica particles of solution pH 6 in our study would be aggregated, hence, particle counts were less than that of the dip solution with pH 7. The solubility of silica as a function of pH is given in Ref. [11] in which it was found that the solubility of silica increases when pH is greater than 8. These silica particles exist in the basic solution in the monomeric form, primarily as Si(OH) 4. Conse- quently, with increasing pH, more particles on the film surface would be desorbed into the solution resulting in lower particle counts.

20

A Sluny porllcle ~ ... O'"" SI3N 4 film

_ o. ~'\ - - ~ sio, mm > E ~ ° ~ t - Q 0. 4 0 q N .6O • i - i -80 - ; 6 ; 10 12 pH

Fig. 1. Change of zeta potential with pH in SC- 1 slurry, SiO 2 [8] and Si3N 4 [9] film.

12000- 10000- o m "--0"-- SIGN4 film

g

Fig. 2. Particle numbers on SiO 2 and Si3N 4 films immersed in various pH solutions for 30 rain after CMP process.

The preceding results can also be perceived as follows [12]: (i) The fresh surface is obtained just after the CMP process is finished. Thus, there is an attraction between particles with negative zeta poten- tial close to the surface and positive ions present in the oxide or bulk of the silicon substrate. This results in a quick particle deposition. (ii) As the wafer is transported to a tank with an alkali solution, the fresh surface gradually disappears and a negative charge (negative zeta potential) is built up on the wafer surface by OH- radicals adsorbed on the surface. So the repulsive forces dominate the interaction between the particle and the surface, leading to weaker con- tact between the two, and chemisorbed silica parti- cles were converted to water soluble residue, which is aqueously removed. This phenomenon will be- come more pronounced when the numbers of OH- radicals or pH value of the dip solution increase.

The zeta potential data in Fig. 1 revealed that Si3N 4 and SiO 2 films have similar global chemical nature, conforming to results from other studies [8,9]. Nevertheless, the deposited silica particle counts of S i a N 4 a r e m u c h l e s s t h a n t h a t o f S i O 2 a s carl b e s e e n in Fig. 2. This implies that contributors other than the electrostatic forces are also responsible for parti- cle deposition or removal in the post-cleaning pro- cess.

In the CMP process, silica particles are fed into a gap between pad surface and polished film. Kragel- sky et al. [3] proposed the probability of adherence

C.-W. Liu et aL /Applied Surface Science 92 (1996) 176-179 179 of particles to pad surface (hardness H 1) and pol-

ished film (hardness H2) are H2 Pr(pad surface) = H1 + H 2 ' H1 Pr(polished f i l m ) = H1 + H 2 " Since H 1 ,t: H2 H1 Pr(polished film)-- H2

This suggests that the number of silica particles physically embedded in the wafer surface is in- versely proportional to the hardness of the polished film, i.e. particles are more difficult to adhere to, or embedded in, the harder surface. Since the hardness of the SiO2(36 GPa) film is less than that of the Si3N 4 film (50 GPa), there are more silica particles counted in the SiO 2 film than in the SiaN 4 film. Most of these debris can be removed via scrubbing process.

4. Conclusion

This study was undertaken to investigate the mechanism of the particle removal in a dip solution after CMP process. The zeta potential of silica parti- cles in solutions with various pH values was deter- mined. Experimental results showed that both the zeta potential of silica particles and silica particle counts decrease as the solution pH increases. This means that the zeta potential is one determining factor for particle deposition after CMP process.

From a statistical viewpoint, particle adherence on the wafer surface is inversely proportional to the

surface hardness of the polished film, therefore, more silica particles were detected on the softer film (SiO 2) than on the harder film (SiaN4).

Acknowledgements

The authors express appreciation for helpful dis- cussions with Dr. T.Y. Yen. This work is supported by National Science Council (NSC84-2622-E009- 007). We thank the Precision Instrument Develop- ment Center for providing the Nanotest 500 used for these experiments.

References

[1] P. Singer, Semicond. lnL (March 1994) 48.

[2] R. Jairath, J. Farlcas, C.K. Huang, M. Stell and S.M. Tzeng, Solid State Technoi. (July 1994) 71.

[3] I.V. Kragelsky, M.N. Dobychin and V.S. Kombalov, Friction and Wear, Calculation Methods (Pergamon, New York,

1982).

[4] L. Mouche, F. Tardif and J. Derrien, J. Electrochem. Soc. 41 (1994) 1684.

[5] W. Kern, Handbook of Semiconductor Wafer Cleaning Tech- nology (Noyes, Park Ridge, NJ, 1993).

[6] R.J. Hunter, Zeta Potential in Colloid Science, Principle and Application (Academic Press, New York, 1981).

[7] M.F. Deemer and W.D. Nix, J. Mater. Res. 1 (1986) 601. [8] L. Bousse, S. Mostarshod, B. Van der Schoot, N.F. de Rooij,

P. Gimme and W. Gopel, J. Coll. Inteff. Sci. 147 (1991) 22. [9] L. Bousse and S. Mostarshed, J. Electroanal. Chem. 302

(199l) 269.

[10] M. Desai, R. Jalrath, M. Stell and R. Tolles, presented at 1994 Materials Research Society Spring Meeting, Boston, MA, USA.

[11] R.K. Iller, The Chemistry of Silica (Wiley-Interscience, New York, 1979) p. 366.