chemical mechanical planarization of barrier layers by using a weakly alkaline slurry
TRANSCRIPT
1
2
3
4 Q1
56
789
1 1
1213 14 15 16
1718 19 20 21 22 23
2 4
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Microelectronic Engineering xxx (2013) xxx–xxx
MEE 9075 No. of Pages 5, Model 5G
13 April 2013
Contents lists available at SciVerse ScienceDi rect
Microe lectronic Engin eering
journal homepage: www.elsevier .com/locate /mee
Chemical mechanical planarization of barrier layers by using a weakly alkaline slurry
Chenwei Wang a,⇑, Jiaojiao Gao a, Jianying Tian b, Xinhuan Niu a, Yuling Liu a
a Institute of Microelectronics, Hebei University of Technology, Tianjin 300130, China b Market Information Department of CSPC Zhongqi Pharmaceutical Technology (Shijianzhuang) CO., Ltd, Shijiazhuang 050051, China
a r t i c l e i n f o a b s t r a c t
25262728293031323334
Article history: Received 8 December 2012 Received in revised form 9 March 2013 Accepted 1 April 2013 Available online xxxx
Keywords:Chemical mechanical planarization Barrier layer Weakly alkaline slurry Dishing and erosion Correction
0167-9317/$ - see front matter � 2013 Published by http://dx.doi.org/10.1016/j.mee.2013.04.001
⇑ Corresponding author. Tel.: +86 022 60204914; faE-mail address: [email protected] (C. Wang).
Please cite this article in press as: C. Wang et a
A weakly alkaline barrier slurry (pH = 8.0) was proposed, which was free of unstable H2O2 and inhibitor such as benzotriazole (BTA). The polishing results of Cu, Ta and oxide blanket wafers show that the slurry has a high removal rate on oxide, while Ta has a low removal rate on Cu. The evaluation of the slurry was implemented in a way that the process conditions of Cu CMP and consumables on platen 1 and platen 2were fixed. Copper dishing and oxide erosion have been characterized as a function of polishing time. The experiment results reveal that the barrier slurry without inhibitors has an obvious effect on the correc- tion of dishing and erosion, and it also suggests that the slurry has a high selectivity of Ta and oxide to Cu. The sheet resistance does not exhibit any difference as polishing time increased, which indicates substan- tially lower copper loss.
� 2013 Published by Elsevier B.V.
35
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
1. Introduction
Barrier layers, such as Ta and/or tantalum nitrides, which are thermodyna mically stable and have low film resistivity have been introduced in integrated circuit interconnec ts (ICs) to prevent the diffusion of Cu atoms into dielectrics and improve the adhesion property between the two layers [1–2]. Due to the incorporation of low-k dielectric s in device structures underlying Ta/TaN barrier layers, it is necessary for chemical mechanical planarizati on (CMP)of barrier at a reduced pressure (61 psi) to prevent structural dam- age to fragile dielectric materials with low-permittivity . In addi- tion, non/weakly alkaline slurry solutions are preferable for CMP of barrier, because many of Si-based low-k materials chemically disintegrate in high pH media [3–5]. During CMP of patterned cop- per wafers, two phenomena – copper dishing and dielectric erosion occur during over-polish step (which is required to ensure com- plete copper removal across the entire wafer).
In this paper, dishing is defined as the recessed height of a cop- per line compared to the neighboring oxide, and erosion is definedas the amount of recess of oxide relative to field oxide surface (Asshown in Fig. 1). Dishing reduces the thickness of wide copper fea- tures, leading to an increase in the resistance and current density along the line. Thus, another major requiremen t for a good barrier slurry is that it can yield high selectivity of TaN: Cu as well as Ta: Cu polish rate to correct the topography [6–10].
Elsevier B.V.
x: +86 022 26556455.
l., Micr oelectron . Eng. (2013), h
Many efforts have been devoted to the developmen t of barrier slurry to improve the CMP performance. However, almost all of these slurries contain H2O2 as oxidizer and benzotriazole (BTA)as inhibitor. As known to all, H2O2 is unstable and easily to be decompo sed, which will deteriorate the performanc e of the slurry. Although BTA has been recognized as an influential inhibitor of copper corrosion in aqueous acidic, neutral, and alkaline solutions. It is reported that BTA creates post-CMP challenges such as leaving hydrophobic copper surface, increasing particle adhesion and increasing etch rate of copper during cleaning [11–17]. Recently, we have reported a kind of barrier slurry without BTA, but it was performed under a relative high pressure (2.0 psi), utilizing unsta- ble H2O2 as oxider [18]. The slurry used for the CMP of barrier is afurther research of our previous investigatio n. The compositi on of the slurry is different from our recent report, key feature of the slurry is free of H2O2 and BTA, slurry performanc es was evaluated in terms of dishing and erosion. Sheet resistance of the copper pat- terned wafer has also been studied, Finally the mechanism of bar- rier CMP was also discussed.
84
85
86
2. Experimen tal
Experime nt was carried out on Applied Materials Reflexion LK 300 mm tool which was designed for a three-step CMP approach .Real-tim e profile control (RTPC) was implemented on platen 1(P1) to provide average removal rate fluctuations and the endpoint of step one at target thickness regardles s of incoming copper film.Subseque ntly, copper clearing was complete d using full scan
ttp ://dx.doi. org/10.1016/ j.mee.201 3.04.001
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
Q2
2 C. Wang et al. / Microelectronic Engineering xxx (2013) xxx–xxx
MEE 9075 No. of Pages 5, Model 5G
13 April 2013
endpoint (EP) on platen 2 followed by barrier removal within the 3rd step. The purpose of this work is to evaluate the performanc eof the barrier slurry, thus the process conditions and consumables on platen 1 and platen 2 were fixed. The Dow chemical IC 1010TM pads with window were used on P1 and P2, Politex Regular em- bossed pad was used for barrier removal. A 15 min pad break-in procedure was performed to condition the pad before experiments. Process parameters in the polishing experime nts are summarized in Table 1. Removal rate selectivity of different blanket wafers
Fig. 1. Schematic representation of cop
Table 1Process parameters in polishing experim ents.
P1
Purpose Bulk copper removal RR/Z 1–Z 5 pressure (psi)a 3.2/2.0/1.5/1.5/1.5/1.5 Slurry flow rate (ml/min) 300 Head rotation speed (rpm) 97 Platen rotation speed (rpm) 103 Endpoint �2 KÅ Cu remaining Conditioner 3 M disc A160 pad conditioner
Fig. 2. The schematic of patterned copp
Fig. 3. Schematic illustration of co
Please cite this article in press as: C. Wang et al., Micro electron. Eng. (2013), h
(including Cu, Ta and Oxide) was evaluated under the same condi- tion of barrier CMP. Schematic of patterned copper wafer was shown in Fig. 2. Copper slurry used on P1 and P2 is a product of Tianjin Jingling Microelectr onic Material limited. Barrier slurry un- der investiga tion was formulat ed with 10 wt.% diluted colloidal sil- ica (20–30 nm in size), 1.5 wt.% FA/O chelating agent (it was obtained from Hebei University of Technolo gy), 0.5 wt.% guanidine nitrate and deionized water (DIW). The pH of the slurry was then adjusted to 8.0 by using phosphoric acid. A RESMAP 463 FOUP
per dishing and dielectric erosion.
P2 P3
Copper clearing Barrier removal 3.2/2.0/1.5/1.5/1.5/1.5 2.2/1.5/1.0/1.0/1.0/1.0 300 300 97 78 103 80 EP 45 s + OP 30 s By time
er wafer used in this experiment.
pper CMP in this experiment.
ttp ://dx.doi. org/10.1016/ j.mee.201 3.04.001
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
C. Wang et al. / Microelectronic Engineering xxx (2013) xxx–xxx 3
MEE 9075 No. of Pages 5, Model 5G
13 April 2013
(Cornell Dubilier Electronics, Inc.) resistivity measure ment was used to measure the copper and Tantalum film thickness. Oxide film thickness was measured by Opti-Prob e 7341 (Therma-WaveInc). The material removal rate of Cu, Ta and oxide was determined by calculating the film thickness before and after polish 60 s. ADimension AFP (Atomic Force Profiler, Veeco) was performed to measure dishing and erosion.
136
137
138
139
140
141
142
143
144
145
146
147
148
Fig. 4. Optical microscope image of CMP pad test structure for measuring dishing and erosion.
3. Results and discussion
Fig. 3 shows the schematic illustration of copper CMP in this experiment. In this study a 3-step polishing scheme was used: During first step, removing most copper, leaving approximat ely 2000 Å and the surface is initially planarized. Step two, removing the remaining copper (always required a short of over-polish step)and stop on barrier layers. Step three, clean barrier metal and some dielectric. The evaluation of the barrier slurry was carried out with different polishing times for 30 s, 40 s, 50 s and 60 s, respectively .
Fig. 4 shows optical microscop e image of monitor pad on copper patterned wafer for measuring dishing and erosion. The monitor pad comprises dishing (70 � 50 lm2 copper pad) and erosion (70 � 50 lm2 array of 140 nm copper lines and 130 nm spaces)features. A short profilometry scan (scan length = 180 lm) across the test pad is made to measure dishing and erosion, and typical examples of atomic force profiler (AFP) profiles are shown in Fig. 5.
Figs. 6 and 7 show the mean values of dishing and erosion along with standard deviation as a function of polishing time. We can see that dishing and erosion values decreased substantially with the increasing polishing time. Dishing values of CMP structure was dramatically reduced from 850 Å to 230 Å after polishing 60 s. Meanwhile, the erosion values decrease d from 358 Å to a low value of 42 Å. The decrease in the dishing values with increased polishing time indicates that alkaline barrier slurry without inhibitors has an
Fig. 5. Typical samples of AFP pro
Please cite this article in press as: C. Wang et al., Micr oelectron . Eng. (2013), h
obvious effect on surface topography correction, it also suggests the slurry has a high removal rate selectivity of barrier layers (Ta/TaN) and cap layers (TEOS) to copper wirings.
Fig. 8 shows the removal rate of Cu, Ta and oxide in weakly alkaline barrier slurry. Because barrier slurry was absence of oxi- dant (such as hydrogen peroxide), oxidation reaction of copper was very weak, and copper surface is mainly compose d of cuprous oxides. Moreover, due to limited ionization of copper, few or no complexi ng reactions occurred between the slurry and copper sur- face. All above factors result in a lower copper removal rate during CMP, the removal rate of copper is about 136 Å/min, consequently the slurry can maintain a low etch rate of copper in the trench, it is helpful for obtaining an effective planarization.
files: (a) dishing, (b) erosion.
ttp ://dx.doi. org/10.1016/ j.mee.201 3.04.001
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
Fig. 6. The mean values of dishing along with standard deviation as a function of polishing time.
Fig. 7. The mean values of erosion along with standard deviation as a function of polishing time.
Fig. 8. Removal rates of Cu, Ta and oxide.
4 C. Wang et al. / Microelectronic Engineering xxx (2013) xxx–xxx
MEE 9075 No. of Pages 5, Model 5G
13 April 2013
Please cite this article in press as: C. Wang et al., Micro electron. Eng. (2013), h
As shows in Fig. 8, Ta removal rate is about 4 times higher than that of Cu, according to the literature, both Ta and TaN can be oxi- dized to Ta 2O5 by O2, atmospheric O2 dissolved in aqueous slurry would play an active role in supporting the reactions. Subse- quently, soluble Ta complexes were produced through direct reac- tions of Ta 2O5 with the slurry, in this case (the pH of the slurry is about 8.0), the soluble Ta complex could be the tetraperoxo tanta- late anion, [Ta(O2)4]3�. Besides, dissolved Ta ions chelated with chelating agent also promote d the removal of the barrier layer, the soluble/weak soluble product could be carried away with flow-ing slurry in CMP [19–21]. For polishing oxide, the maximum re- moval rate of oxide is 1056 Å/min. The result shows that the oxide removal rate is significantly higher than that of Cu. The high selectivity of oxide to Cu is effectively to minimize dishing and ero- sion. The polishing mechanism of the oxide can be explained by that the oxide surface may react with the hydroxyl ions in the weakly alkaline solution and form soluble silicate. The hydrolysis reaction can be represented by the following equation:
167
169 169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
SiO2 þ 2OH� ! SiO2�3 þH2O
In the ideal case, after the removal of barrier layers, the exposed oxide layer (TEOS) should be complete ly removed and stop on the low-K dielectric. Based on the results of dishing and erosion, although the barrier slurry has a good performance for topograp hy correctio n, it also has removal rate on low-k dielectric, and we will solve the problem in further research. In order to prevent more loss of low-k dielectric, the end point for barrier polishing is 40 s.
Fig. 9 presents sheet resistance (Rs) of copper patterned wafer on 0.1 lm line width (Lw) and 0.1 lm line space (Ls) as a function of polishing time. In general, sheet resistance will increase as polishing time increase. Most interestingly , it can be seen from Fig. 9, sheet resistance doesn’t show any difference as polishing time increses, indicating substantially lower copper loss. The samples polished 30 s give an Rs result of 0.155 X/sq with a tight standard deviation (Std Dev) of 0.009 X and the sample polished
Fig. 9. Sheet resistance values of copper patterned wafer on Lw/Ls = 0.1 lm/0.1 lmas a function of polishing time.
Table 2Mean and standard deviation of sheet resistance as a function of polishing time.
Polishing time (s) Mean (X/sq) Std dev (X/sq)
30 0.155 0.009 40 0.157 0.008 50 0.155 0.007 60 0.159 0.007
ttp ://dx.doi. org/10.1016/ j.mee.201 3.04.001
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206 207 208 209 210 211 212 213 214 215 216 217
C. Wang et al. / Microelectronic Engineering xxx (2013) xxx–xxx 5
MEE 9075 No. of Pages 5, Model 5G
13 April 2013
60 s also obtains the same Rs of 0.155 X/sq with lower standard deviation of 0.007 X as shown in Table 2. We can see that the slur- ry without inhibitor can protect the copper from directly corrosion and has a low removal rate on copper, leading to lower copper loss. All these probably due to the insoluble oxide and hydroxide of cop- per formed on the lower region in the trench. These results coin- cide well with previous observati on of correction effect of dishing and erosion, attributing from the formatio n of protect film,which is also a major reason for fewer copper loss. Additionally ,within wafer non-uniform ity was also reduced as the polishing time increased, having a narrow Rs derivation.
218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236
4. Conclusion
A weakly alkaline barrier slurry (pH = 8.0) was proposed , which was free of unstable H2O2 and inhibitor such as benzotriazole (BTA).The experiment results reveal that the alkaline barrier slurry without inhibitors has an effectively effect on the correction of surface topog- raphy, it also suggests that the slurry has a high removal rate selectiv- ity of barrier layers (Ta/TaN) and cap layers (TEOS) to copper wirings. The sheet resistance does not exhibit any difference as polishing time increased, which indicates substanti ally lower copper loss.
237
Please cite this article in press as: C. Wang et al., Micr oelectron . Eng. (2013), h
References
[1] Y. Li, S.V. Babu, Electrochem. Solid State Lett. 4 (2001) G20. [2] P.B. Zantye, A. Kumar, A.K. Sikdar, Mater. Sci. Eng. R Rep. 45 (2004) 89. [3] S.E. Rock, D.J. Crain, C.M. Pettit, D. Roy, Thin Solid Films 520 (2012) 2892–2900.[4] S.E. Rock, D.J. Crain, J.P. Zheng, C.M. Pettit, D. Roy, Mater. Chem. Phys. 129
(2011) 1159–1170.[5] C.M. Sulyma, D. Roy, Corros. Sci. 52 (2010) 3086–3098.[6] S.V.S.B. Janjam, B.C. Peethala, J.P. Zheng, S.V. Babu, D. Roy, Mater. Chem. Phys.
123 (2010) 521–528.[7] C.M. Sulyma, D. Roy, Appl. Surf. Sci. 256 (2010) 2583–2596.[8] K. Gottfried, I. Schubert, S.E. Schulz, T. Gessner, Microelectron. Eng. 83 (2006)
2218–2224.[9] C.M. Pettit, D. Roy, Mater. Lett. 59 (2005) 3885–3889.
[10] Y.H. Chen, T.H. Tsai, S.C. Yen, Microelectron. Eng. 87 (2010) 174–179.[11] S.Y. Chiu, Y.L. Wang, C.P. Liu, S.C. Chang, G.J. Hwang, M.S. Feng, C.F. Chen, Thin
Solid Films 498 (2006) 60–63.[12] M. Nagar, J. Vaes, Y. Ein-Eli, Electrochim. Acta 55 (2010) 2810–2816.[13] T. Koito, K. Hirano, K. Nakabeppu, IEEE Trans. Semi. Manufac. 15 (2002) 429. [14] P.L. Chen, J.H. Chen, M.S. Tsai, B.T. Dai, C.F. Yeh, Microelectron. Eng. 75 (2004)
352. [15] J.W. Lee, M.-C. Kang, J.J. Kim, J. Electrochem. Soc. 152 (2005) C827. [16] K. Holloway, P.M. Fryer, C. Cabral Jr., J.M.E. Harper, P.J. Bailey, K.H. Kelleher,
J. Appl. Phys. 71 (11) (1992) 5433–5444.[17] K.H. Min, K.C. Chun, K.B. Kim, J. Vac. Sci. Technol. B 14 (5) (1996) 3263–3269.[18] C.W. Wang, Y.L. Liu, J.Y. Tian, B.H. Gao, X.H. Niu, Microelectron. Eng. 98 (2012)
29–33.[19] S.V.S.B. Janjam, B.C. Peethala, D. Roy, S.V. Babu, Electrochem. Solid State Lett.
13 (1) (2010) H1–H4.[20] S.V.S.B. Janjam, C.V.V.S. Surisetty, S. Pandija, D. Roy, S.V. Babu, Electrochem.
Solid State Lett. 11 (3) (2008) H66–H69.[21] S.V.S.B. Janjam, S. Peddeti, D. Roy, S.V. Babu, Electrochem. Solid State Lett. 11
(12) (2008) H327–H330.
ttp ://dx.doi. org/10.1016/ j.mee.201 3.04.001