Microelectronic Engineering 87 (2010) 370–372
Contents lists available at ScienceDirect
Microelectronic Engineering journal homepage: www.elsevier.com/locate/mee
Effect of ﬂuorine contamination on barrier metal oxidation S. Ozaki a,*, Y. Nakata a, Y. Kobayashi a, T. Nakamura a, Y. Iba b, S. Fukuyama b, H. Watatani b, Y. Ohkura b a b
Fujitsu Laboratories Ltd., 10-1 Morinosato-Wakamiya, Atsugi, Kanagawa 243-0197, Japan Fujitsu Microelectronics Ltd., 1500 Mizono, Tado-cho, Kuwana, Mie 511-0192, Japan
a r t i c l e
i n f o
Article history: Received 24 March 2009 Received in revised form 10 June 2009 Accepted 22 June 2009 Available online 25 June 2009 Keywords: Barrier metal Oxidation Interfacial Fluorine contamination Porous Low-k Etching
a b s t r a c t We clariﬁed that interfacial barrier metal oxidation with inter layer dielectric (ILD) could be revealed by X-ray photoelectron spectroscopy (XPS) on the peeled-side of the barrier metal, and the barrier metal oxidation was promoted by ﬂuorine contamination which adsorb to the ILD surface during etching. To consider the effect of ﬂuorine contamination on barrier metal oxidation, hydrolysable property of ﬂuorine contamination was evaluated by measuring the change of F 1s spectrum after dipping in boiling water. Moreover, ﬂuoride ions and the acidity of water in which ﬂuorine contamination was dipped were measured by Ion chromatography and pH measurement, respectively. According to our experiments, it was suggested that hydroﬂuoric acid (HF) acted as an oxidizing catalyst to promote barrier metal oxidation at the interface of barrier metal and ILD. Ó 2009 Elsevier B.V. All rights reserved.
Minimizing the resistance–capacitance delay of interconnects is important for making high-speed devices, and this is driving the need for the production of porous low-k ILD. However, these materials have several difﬁculties when integrated into Cu interconnects. One of the major problems with integrating porous low-k ILD is the plasma-induced damage caused by dry etching. This is because the change in ﬁlm characteristics induced by the plasma degrades interconnects characteristics and reliability. It is well known that the barrier metal is oxidized because of moisture which adsorb to the porous low-k ILD by plasma-induced damage, and the barrier metal oxidation is considered one of the possible causes for stress-induced voiding (SIV) [1,2]. Therefore, when the barrier metal thickness decreases with CMOS node shrinkage, the barrier metal oxidation would become much more serious. On the other hand, some studies have proved that ﬂuorine contamination adsorb to the ILD surface during etching [3,4]. However, no work has reported on the effect of ﬂuorine contamination on barrier metal oxidation. If ﬂuorine contamination reacts with moisture which absorb to the ILD, the acidity of moisture would be changed, and the barrier metal oxidation would be promoted. In this study, we investigated the effect of ﬂuorine contamination on barrier metal oxidation.
The ILD referred to in this paper was silica-based spin-on dielectrics (SOD) with a dielectric constant of less than 2.3 . Samples-A and B were 150-nm thick ILD ﬁlms deposited on bare Si wafer, and only Sample-B was etched using a conventional 13.56 MHz capacitively coupled plasma (CCP) etcher with CF4 chemistry. Chemical changes due to etching were evaluated by X-ray photoelectron spectroscopy (XPS). Hydrolysable property of ﬂuorine contamination which adsorbed to the ILD surface was evaluated by measuring the change of F 1s spectrum after dipping Sample-B in boiling water for 1 h. Fluoride ions and the acidity of water in which Sample-B was dipped were measured by Ion chromatography and pH measurement, respectively. Change in the corrosive property of water caused due to dipping of Sample-B was evaluated by measuring corrosion current. Ta, Pt and Ag/AgCl were used as working electrode (WE), counter electrode (CE) and reference electrode (RE), respectively. The barrier metal oxidation was evaluated for the samples with Cu/Ta/ILD/Si substrate structure. The barrier metal of 30 nm thick Ta layer was deposited on Samples-A and B in Radio Frequency (RF) sputtering system, and 50 nm thick Cu layer was subsequently deposited on the Ta ﬁlm in the same sputtering system. The Cu/Ta ﬁlm was peeled from the ILD/Si substrate under a nitrogen atmosphere condition in a glove box, and the samples were transferred to XPS load lock chamber using a specimen transfer box ﬁlled with nitrogen gas in order to prevent the specimen from exposed to the air. Ta 4f spectrum on the peeled-side was measured by angle-re-
* Corresponding author. Tel.: +81 46 250 8262; fax: +81 46 250 8235. E-mail address: [email protected]
(S. Ozaki). 0167-9317/$ - see front matter Ó 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.mee.2009.06.011
S. Ozaki et al. / Microelectronic Engineering 87 (2010) 370–372
solved XPS. The take-off angle ranging from 90° to 30° were used to change the surface sensitivity. The Ta2O5 rate was calculated by a waveform analysis of separating the Ta 4f spectrum into Ta (21.5 and 23.5 eV) and Ta2O5 (26.5 and 28.5 eV). 3. Results and discussion 3.1. Evaluation of chemical changes due to ﬂuorine contamination To evaluate chemical changes of ILD surface due to etching, Samples-A and B were measured by XPS. Fig. 1a and b shows C 1s and F 1s spectrum, respectively. The peaks corresponding to CFx (287–294 eV) and Si–CH3 (284–286 eV) [6,7] were observed in C 1s spectrum as shown in Fig. 1a. It is considered that Si–CH3 which observed in both samples is regarded as ILD bonds, while CFx is due to the etching process. In addition, SiFx (685–687 eV)  was observed in F 1s spectrum of Sample-B as shown in Fig. 1b. These results indicated that ﬂuorine contamination due to etching existed as CFx and SiFx. According to Kajihara et al. SiFx spontaneously reacts with water, thus, hydrolysis and polycondensation of SiFx in oxyhydrogen ﬂame or aqueous solutions are widely used to fabricate SiO2 . So, there is a possibility that SiFx reacts with moisture which absorbs to the ILD, and the acidity of moisture would be changed. To evaluate hydrolysable property of ﬂuorine contamination, the change of F 1s spectrum before and after dipping Sample-B in the boiling water were measured by XPS. Fig. 2 shows the F 1s spectrum of the Sample-B before and after dipping. The decrease of SiFx observed in F 1s spectrum was attributed to the hydrolysis reaction. On the other hand, the decrease of CFx was not observed
Fig. 2. Change of SiFx peak intensity before and after Sample-B (after etching) dipping in water.
in C 1s spectrum because of their hydrophobic property. To evaluate ﬂuoride ions and the acidity of water in which Sample-B was dipped, Ion chromatography and pH measurement were conducted. As shown in Table 1, ﬂuoride ions were detected in water after dipping Sample-B, and a pH decreased with increase of ﬂuoride ions concentration. This tendency is reasonable from the perspective of solution chemistry. Therefore, it is considered that SiFx reacted with water to form HF as follows :
BSi—F þ H2 O ! BSi—OH þ HF ð3—1Þ 2BSi—F þ H2 O ! —Si—O—SiB þ 2HF ð3—2Þ To evaluate the corrosive property of water which was used for the dipping experiments, corrosion current of Ta was measured. Fig. 3 shows the changes in the corrosive current caused by HF generation. The increase of corrosion current was observed after dip-
Table 1 Fluoride ions concentration and pH of water before and after dipping Sample-B (after etching).
Fig. 1. XPS spectrum of ILD surface before and after etching: (a) C 1s; (b) F 1s.
Fluoride ion concentration [ppb]
Before dipping Sample-B After dipping Sample-B
<5 (below detection limit) 240
Fig. 3. Change of corrosion current before and after Sample-B (after etching) dipping in water.
S. Ozaki et al. / Microelectronic Engineering 87 (2010) 370–372 Table 2 Ta2O5 rate which calculated by waveform analysis of Ta 4f spectrum. Sample name
Rate (%) 90° take-off angle
Sample-A (before etching) Sample-B (after etching)
30° take-off angle
3–4 nm for the take-off angles of 90° and 30°, respectively. Therefore, it is considered that Ta2O5 layer was formed with a thickness of less than 3–4 nm at the interface of the barrier Ta and ILD. To evaluate Ta2O5 rate, waveform analysis was performed by separating the Ta 4f spectrum into Ta and Ta2O5. As shown in Table 2, the Ta2O5 rate increased due to etching. Moreover, the increase of the Ta2O5 rate was larger in 30°take-off angle than in 90° take-off angle. This result strongly indicated that the barrier metal oxidation was promoted by ﬂuorine contamination which was present at the Ta/ILD interface. Therefore, it is considered that hydrolysable SiFx reacted with water to form HF and this acted as an oxidizing catalyst of Ta oxidation. 4. Conclusion
Fig. 4. Ta 4f spectrum on the peeled-side of Ta before and after etching: (a) 90° take-off angle; (b) 30° take-off angle.
ping of Sample-B. So, it is considered that HF changed the acidity of water and increase corrosion current of Ta. Therefore, if a similar reaction occurs at the Ta/ILD interface, HF would act as an oxidizing catalyst to promote Ta oxidation. 3.2. Evaluation of barrier metal oxidation To evaluate interfacial Ta oxidation, the Cu/Ta layer was peeled from the ILD/Si substrate without exposed to the air and subsequently analyzed the peeled-side of Ta by angle-resolved XPS. Fig. 4a and b shows the Ta 4f spectrum which measured by 90°and 30° take-off angle, respectively. Ta (21.5 and 23.5 eV) and Ta2O5 (26.5 and 28.5 eV)  were observed in both ﬁgures. This means that interfacial Ta oxidation with ILD could be detected by means of XPS analysis on the peeled-side of Ta. The effect of the ILD surface etching on the Ta oxidation was compared in each ﬁgure. In the conﬁguration of 90° take-off angle, the peak intensity of Ta2O5 in the Sample-B was larger than in the Sample-A while the difference of the Ta peak intensity was not so large. On the other hand, the Ta peak intensity in the Sample-A was larger than that in the Sample-B in the 30° take-off angle conﬁguration as shown in Fig. 4b. The detection depths were 7–8 nm and
We clariﬁed that interfacial barrier metal oxidation with ILD could be revealed by XPS measurements on the peeled-side of barrier metal, and barrier metal oxidation was promoted by ﬂuorine contamination which adsorb to ILD surface during etching. Fluorine contamination existed as CFx and SiFx, and hydrolysable SiFx reacted with water to form HF and increased corrosion current of water. These results strongly indicated that HF acted as an oxidizing catalyst to promote barrier metal oxidation at the interface of barrier metal and ILD. Similarly, any ﬂuorine contamination of actual Cu interconnects would promote the barrier metal oxidation. References  N. Matsunaga, N. Nakamura, K. Higashi, H. Yamaguchi, T. Watanabe, K. Akiyama, S. Nakao, K. Fujita, H. Miyajima, S. Omoto, A. Sakata, T. Katata, Y. kagawa, H. Kawashima, Y. Enomoto, Proc. IITC (2005) 6–8.  A. Sakata, S. Yamashita, S. Omoto, M. Hatano, J. Wada, K. Higashi, H. Yamaguchi, T. Yosho, K. Imamizu, M. Yamada, M. Hasunuma, S. Takahashi, A. Yamada, T. Hasegawa, H. Kaneko, Proc. IITC (2006) 101–103.  Y. Iba, T. Kirimura, M. Sasaki, Y. Kobayashi, Y. Nakata, M. Nakaishi, Jpn. J. Appl. Phys. 47 (2008) 6923.  F. Furukawa, R. Wolters, H. Roosen, J.H.M. Snijders, R. Hoofman, Microelectron. Eng. 76 (2004) 25.  I. Sugiura, N. Misawa, S. Otsuka, N. Nishikawa, Y. Iba, F. Sugimoto, Y. Setta, H. Sakai, Y. Koura, K. Nakano, T. Karasawa, Y. Ohkura, T. Kouno, H. Watatani, Y. Nakata, Y. Mizushima, T. Suzuki, H. Kitada, N. Shimizu, S. Nakai, M. Nakaishi, S. Fukuyama, T. Nakamura, E. Yano, M. Miyajima, K. Watanabe, Microelectron. Eng. 82 (2005) 380.  C. Huang, Y. Wang, S. Chang, G. Hwang, J. Huang, Thin Solid Films 498 (2006) 286–288.  M. Touzin, P. Chevallier, F. Lewis, S. Turgeon, S. Holvoet, G. Laroche, D. Mantovani, Surf. Coat. Technol. 202 (2008) 4884–4891.  S. Gao, M. Lei, Y. Liu, L. Wen, Appl. Surf. Sci. 255 (2009) 6017–6023.  K. Kajihara, M. Hirano, L. Skuja, H. Hosono, J. Non-Cryst. Solids 353 (2007) 514– 517.  C. Xu, H. Dong, L. Yuan, H. He, J. Shao, Z. Fan, Opt. Laser Technol. 41 (2009) 258–263.