January 26, 2007

The Evolution of Diffusion Barriers in Copper Metallization

By Lee, Chiapyng; Kuo, Yu-Lin

Refractory metal nitride thin films have been widely developed as the diffusion barriers for the aluminum or copper interconnects in integrated circuits. This study reviewed the evolution of diffusion barriers in copper metallization. First, materials characteristics and electrical properties of varions diffusion barriers, titanium nitride (TiN), tantalum nitride (TaN), and titanium zirconium nitride (TiZrN), were examined. These diffusion barriers were prepared by reactive magnetron sputtering in N^sub 2^/Ar gas mixtres. Next, barrier performance was evaluated by annealing the Cu/ barrier/Si systems at 400-1,000C for 60 min. in vacuum as well as the measurements of copper diffusion coefficients. The results suggest that TiZrN films can be used as a diffusion barrier for copper metallization better than the well-known TaN films. Therefore, the evolution of diffusion barriers in copper metallization, from TiN to TaN and then from TaN to TiZrN, is addressed.


Melallization is a well-developed technology for the fabrication of integrated circuits based on silicon substrates, as represented in Figure 1a. Owing to the projection of miniaturization of microclecironic devices made by the International Technology Roadmap for Semiconductors,1 the interconnect technology in 2009 evolves toward the 50 run node. Copper is a promising interconnect material for the ultra-large semiconductor integrated circuit (ULSI) due to its lower electrical resistivity (bulk, 1.67 ohm-cm) and higher resistance against electromigration than aluminum and its alloys.2 However, the interaction between copper and silicon is severe and detrimental to the electrical performance of silicon ULSIs even at temperatures as low as 200C. In addition, copper diffuses rapidly in silicon or SiO^sub 2^, which also deteriorates the device operation.3 Therefore, it is necessary to insert a barrier between copper and silicon to isolate the copper, as shown in Figure 1b. In addition, future barriers must not only be successful in resisting a copper penetration of 30 but also have good adhesion performance.1 Until now, refractory metal nitrides have been generally know as diffusion barriers owing to their high melting points, high thermal stability, and high conductivity.

Various refractory metal nitride diffusion barriers, such as Ti- N,4-6 W-N,7 Zr-N,8 Ta-N,9-11 and TiZr-N12-15 have been widely employed in copper metallization. T. Oku et al.11 firstly examined the evolution of diffusion barriers in copper metallization. They plotted the thicknesses of various diffusion barriers vs. the highest annealing temperatures at which the diffusion barriers of known thickness were penetrated by copper. According to their study,11 the future research direction is that thinner diffusion barriers are employed in metallization and required to be stable at sufficiently high temperatures. In addition, C.K. Hu et al.16 developed a relationship to describe the effects of resistivity and thickness of diffusion barriers on the effective resistivity of copper interconnects. This relationship can be expressed by Equation 1 (all equations are shown in the Equations table), where M^sub 1^ and M^sub 2^ are the width and height of metal interconnects, B is the thickness of diffusion barriers, and ρ^sub b^ is the resistivity of diffusion barriers. Apart from the issue of thickness effect, the resistivity of diffusion barriers apparently influences the effective resistivity of the copper interconnects based on Equation 1. Therefore, a copper interconnect with low effective resistivity is obtained from a diffusion barrier with low resistivity.

In comparison with TaN, TiN is not as effective a barrier against copper diffusion.11 However, TiZrN may be a good candidate for a copper diffusion barrier, due to its low bulk resistivity (TiZr ~ 26 ohm-cm, TiZrN ~ 60 ohm-cm) and excellent adhesion to copper. Additionally, the use of TiN barriers in Al/Si metallization is mature. Thus, the applicability of TiZrN to copper-based interconnects is obvious. This review compares the performances of the three well-know refractory metal nitrides: titanium nitride (TiN), tantalum nitride (TaN), and titanium zirconium nitride (TiZrN), employed as diffusion barriers in Cu/barrier/Si systems. See the sidebar for experimental methods.


Material Characteristics of Sputtered Diffusion Barriers

Three metal nitride (TiN, TaN, and TiZrN) films used as diffusion barriers were reactively sputtered on silicon substrates in N^sub 2^/ Ar gas mixtures. Figure 2 discloses their possession of face- centered cubic (fee) structure tor the TiN turn with (111), (200), and (220) peaks, TiZrN film with a (111) and (200) peaks, andTaN film with a (111) peak appearing in their x-ray diffraction (XRD) spectra, respectively. The differences in the 2θ values of XRD peaks are due to the differences in the lattice constants. Table I shows the lattice constants derived from the XRD patterns. The lattice constants of the deposited TiN and TaN films are 4.234 and 4.323 , respectively, which are slightly smaller than those of standard TiN (4.241 ) and TaN (4.335 ), respectively.17,18 In addition, the lattice constant of TiZrN films (4.303 ) corresponding to an enlargement of 0.062 or 1.46% as compared to that of standard TiN might be due to the incorporation of zirconium atoms, which have a larger atomic radius than that of titanium atoms (titanium: 2.0 , zirconium: 2.16 ) into the lattice. The lattice constant of TiZrN film derived from its XRD pattern is 4.303 .

Figure 3 shows the cross-sectional transmission-electron microscopy (XTEM) bright-field micrographs and electron-diffraction (ED) patterns of these three barrier films. As can be seen in the ligure, the bright-field images of the films show the typical columnar structures. From the indexation of diffraction rings, the microstructures of these barrier films can he described as an assembly of small crystallites with face-centered cubic (fee) structure. The 4.307 lattice constant of TiZrN obtained from its TEM ED pattern as shown in Figure 3c, is consistent with 4.303 derived from its XRD pattern (Figure 2). Therefore, the XRD and TEM results, which evidently show TiZrN films have an fee structure and an enlarged lattice constant as compared to than of TiN lattice, are consistent with the observation of O. Knotek et al.12,13 They also reported that the amount of lattice constant enlargement depends on the content of zirconium atoms. Additionally, they proposed that zirconium atoms substituted randomly the titanium atoms of the TiN lattice and enlarge the lattice constant because of the larger radius of the zirconium atoms as compared to that of titanium atoms.12,13

Refracting metal nitride films are expected to have a defect structure where deviations from stoichiometry are frequent.19 In previous reports,10,20 this phenomenon was verified by the x-ray photoelectron spectroscopy (XPS) studies (not shown here) in which the Ta 41 and Ti 2p peaks in TaN and TiZrN films shifted, respectively, with variations in N^sub 2^/Ar flow ratio and the nitrogen content of the deposited films increased with increasing N^sub 2^/Ar flow ratio. This is expected because nitrogen doping should affect the chemical environment of the outermost electron orbitals of metal atoms. Moreover, A.E. Schmid et al.21 proposed that with sub-stoichiometric compositions (N/Ti

Table I also displays the chemical composition, film density, and film resistivity of the three barrier films. Film densities were calculated from XPS and XRD spectra, as described by W.D. Callister.22 film density of TiZrN barrier films is higher than that of TiN films because of zirconium incorporation into the TiN lattice but smaller than that of TaN films. For the film resistivity of the three deposited barrier films, TiXrN films have a lower film resistivity (60 μOhm-cm) than those of TiN (170 μOhm-cm) and TaN (350 μOhm-cm). It was reported that the required film resistivity of diffusion barriers for ULSI generation has to be lower than 300 μOhm-cm, which means TiZrN films could be good candidates as diffusion barriers in the future.6

Thermal Stability of Cu/Barrier/Si Systems

Twenty-five nanometer TiN, TaN, and TiZrN films reactively sputtered on silicon substrates were employed as diffusion harriers tor Cu/Si metallization. After deposition of a 100nm thick copper film. the Cu/barrier/Si samples were subjected to heat treatment at 400-1,000C for 60 min. in vacuum (10^sup -6^ torr). The variation of copper sheet resistance as a function of the annealing temperature was used to examine the capability of diffusion barrier against copper diffusion. The difference in sheet resistance between the annealed and as-depositedsamples divided by the sheet resistance of as-deposited samples is called the variation percentage of sheet resistance (ΔR^sub s^/R^sub s^%) and is defined in Equation 2.23

It is generally known that Cu-Si compounds are formed at a temperature as low as 200C, and the formation of Cu-Si compounds results in the increase of sheet resistance of copper films in the Cu/barrier/Si samples. Figure 4 presents the variation percentage of sheet resistance versus annealing temperature for the three Cu/ barrier/Si samples. It shows that the sheet resistance of the three Cu/barrier/Si samples remains stable after annealing at temperatures up to 600C for 60 min. However, drastic increases in sheet resistance were found after annealing above 600C for the Cu/TiN/Si sample, 700C tor the Cu/ TaN/Si sample, and 800C for Cu/TiZrN/ Si sample. As a result, TiZrN barrier film has a higher resistance in copper diffusion as compared to those of TiN and TaN films.

To understand the reason for the variation in sheet resistance of Cu/barrier/Si samples after annealing, XRD, plane-view scanning- electron microscopy (SEM), and XTHM were performed to investigate the reactions between the layers. Figure 5a and 5b showed the XRD spectra of Cu/TiZrN/Si and Cu/TaN/Si samples before and after annealing in the temperature range of 400-1,000C for 60 min. in vacuum. In Figure 5a, XRD reveals the diffraction peaks of copper and TiZrN for the as-deposited, and annealed at 700C, 800C. 850C, and 900C Cu/TiZrN/Si samples. Figure 5a shows that at 850C, the Cu (111) peak has sharply decreased in intensity, and subsequently at 900C, diffraction peaks of Cu^sub 3^Si appear at 2θ values of 44.75 and 45.39 which can be assigned as Cu^sub 3^Si (320) and Cu^sub 3^Si (312),24 respectively. Therefore, above 850C the barrier fails. At the same time TiSi^sub 2^ can also he observed upon annealing at 900C. In addition, the disappearance of copper in the XRD pattern (Figure 5a) indicates that most of the copper has diffused through the barrier film to the silicon substrate where it reacts with silicon to form Cu^sub 3^Si after annealing at 900C. Therefore, the formation of high-resistivity Cu^sub 3^Si corresponds to the drastic increase in sheet resistance of copper films, as shown in Figure 4. Similar results of XRD and sheet resistance measurements were also obtained for the Cu/TaN/Si samples shown in Figure 5b except that the appearance of Cu^sub 3^Si and TaSi^sub 2^, and the drastic increase of sheet resistance occurred after annealing at 800C.

Figure 6a and b represents the plane-view SEM and XTEM micrographs of Cu/TiZrN/Si samples after annealing at 900C for 60 inin. In the plane-view SEM and XTEM micrographs, large square pinholes on the ruptured copper films (Figure 6a) and the triangular crystallite with a sharp tip penetrating into the silicon substrate (Figure 6b) were clearly observed. By the use of energy-dispersive analysis of x-rays (EDAX), Figure 7a discloses that the triangular crystallite contains mostly copper and some silicon. In contrast to the XRD observation (Figure 5a), therelore, it is Cu^sub 3^Si. From the value of the angle between the inclined edges as shown in Figure 6b, it is deduced that the crystallites are hounded by the Si {111} planes. A similar pyramidal structure was also observed in the annealed Cu/Ta/Si system, and the pyramid was identified as Cu^sub 3^Si.25

The solid-phase interaction and/or diffusion taking place at the interfaces in the Cu/barrier/Si samples before and after annealing were also examined with an x-ray photoelectron spectroscopy (XPS) depth profile. Figure 7b shows the XPS depth profile of the 900C annealed Cu/TiZrN/Si sample, which demonstrates a complete inlermixing among five elements (Cu, Ti, Zr, N, and Si) between layers in the sample. Therefore, interfaces between layers no longer exist due to the diffusion of copper into silicon substrates through the barrier film and the nut-diffusion of titanium, zirconium, and nitrogen, indicating a catastrophic failure of the Cu/TiZrN/Si sample.

The failure mechanism of the Cu/barrier/Si system has been studied previously by 1. Chen and J.L. Wang.25 The pmposed mechanism is that copper breaks Si-Si bonds as it reaches the TiN/Si interface and silicon point detects generated therein form Cu^sub 3^Si and TiSi^sub 2^. Therefore, it is highly expected that, as copper promotes the formation of silicides, copper may act as a catalyst for the silicidation of the TaN and TiZrN in this study (Figure 5a and 5b). In addition, the concentration of these point defects is greatly enhanced during thermal annealing and promotes the reaction of silicon with barriers, and then the formation of Cu^sub 3^Si and TiSi^sub 2^/laSi^sub 2^. Therefore, the failure of a multiluyered system can he determined when the formation of Cu^sub 3^Si and TiSi^sub 2^/TaSi^sub 2^ (Figure 5a and 5b) and the intermixing phenomena in XPS depth profile (Figure 7b) are observed. This process will proceed until all the copper is consumed.

Oku et al.11 first reported that diffusion coefficients of copper in barrier films can be estimated from the evolution of the XRD profiles as a function of time and temperature, since the diffusion must comply with Fick's first and second laws. They proposed that the average diffusion length of copper atoms (L) can he approximated by 2(Dt)^sup 1/2^, where D is the diffusion coefficient of copper in the barrier and t is the diffusion time.11 In a previous study,15 the authors developed a new and precise methodology to determine the diffusion coefficients of copper in barrier films through the variation of copper sheet resistance as a function of annealing time. With this method,15 the threshold time (t^sub threshold^) for the failure of the barrier, which is the diffusion time, could be determined and then be used in the calculation of the diffusion coefficient of copper in barriers. Figure 8 shows a comparison of copper diffusion coefficients in different diffusion barriers. In Figure 8, diffusion coefficients of copper in TiZrN films are smaller than those of copper in TaN and TiN, which means TiZrN barrier films could withstand the copper diffusion better than TaN and TiN barrier films. Therefore, TiZrN can be used as a diffusion barrier for copper metallization which performs better than TaN, and meets the future research direction of diffusion barriers.


Four-inch n-type Si (100) wafers of resistivity 1-10 ohm-cm were used as substrates in this study. Diffusion barrier films (TiN, TaN, and TiZrN) were reactively sputten from titanium, tantalum, and titanium zirconium Ti-20at.%Zr alloy targets in N^sub 2^/Ar gas mixtures, respectively. The base pressure of the deposition chamber was below 1.3 10^sup -5^ Pa. During the deposition, the total pressure was maintained constant at 0.13 Pa. The radio frequency (RF) power applied to the targets and the argon flow rate were maintained at 250 W and 8 seem, respectively, while the N^sub 2^/Ar flow ratio was adjusted by varying the N^sub 2^ flow rate. Notably, the substrate holder was neither heated nor cooled during deposition, and the substrate bias was held constant at 0 V.

The crystalline structure of the deposited metal nitride barrier films was examine by a Rigaku RTP 300RC x-ray diffractometer (XRD, Cu Kα radiation of λ = 1.5408 ) and cross-sectional transmission electron microscope (XTEM), A JEOL 2000 FXII scanning- transmission-electron microscope with an acceleration voltage of 200 kV was used to examine samples which were prepared in a cross- sectional view. Film thickness was directly measured from the XTEM micrographs. Chemical compositions of barrier films measured by x- ray photoelectron spectroscopy (XPS) analysis were performed in a Thermo VG Scientific Theta Probe spectrometer. All XPS data presented herein were acquired using a monochromatized Al Kα line (hυ = 1,486.6 eV). Peak positions were then calibrated with respect to the C 1s peak at 284.5 eV from the adventitious hydrocarbon contamination. The film resistivity was calculated from the sheet resistance measured by a four-point probe (far field pattern (FPP), Napson RT-7) and the film thickness measured by XTHM.

The thermal stability of the TiN, TaN, and TiZrN diffusion barriers was investigated by employing 25 nm thick nitride films as diffusion harriers for Cu/Si metallization. After deposition of a 100 nm thick copper film, the Cu/barrier/Si samples were subjected to heat treatment at 400-1,000C for 60 min. in vacuum. The pressure for each annealing condition was below 6.7 10^sup -5^ Pa. Variation in sheet resistance of samples, before and after annealing, was measured by a four-point probe. Surface morphology and phase formation of the Cu/barrier/Si samples after annealing were characterized by fieldemission scanning electron microscopy and XRD, respectively. Interfacial behavior was investigated by XTHM. Constituent elements at local regions of the samples were analyzed by energy-dispersive analysis of x-rays. The diffusion coefficients of copper in diffusion barrier films can be estimated from the evolution of the FPP method as a function of time and temperature (500-900C), which was developed and reported in our previous study.15


1. International Technology Roadmap for Semiconductors-2004 Update (Semiconductor Industry Association, 2004), p. 7; http:// public.itrs.net/.

2. J.D. Plummet, M.D. Deal, and P.B. Griffin, Silicon VLSI Technology (Upper Saddle River, NJ: Prentice Hall, 2000), p, 695.

3. Y.S. Diamand el at., Proc. 9 Bienn. Univ. Gov. Ind. Microelectron. Symp. (Piscataway, NJ: IEEE, 1991), pp, 210-215.

4. J.O. Olowolafe, J. Li, and J.W. Mayer, Applied Surface Science, 58 (1991), p,469.

5. Y.S. Gong, J.C. Lin, and C. Lee, Applied Surface Science, 92 (1996), p. 335.

6. M. Moriyama et al., Thin Solid Films, 416 (2002), p. 136.

7\. P.J. Pokela et al., Applied Surface Science, 53 (1991), p. 364.

8. M.B. Takeyama et al., Applied Surface Science, 190 (2002), p. 450.

9. X. Sun et al., Thin Solid Film, 236 (1993), p. 347.

10. J.C. Lin, G. Chen, and C. Lee, J. Electrochem. Soc., 146 (1999), p. 1835.

11. T. Oku et al., Applied Surface Science, 99 (1996), p. 265.

12. O. Knotek, M. Bohmer, and T. Leyendecker, J. Vac. Sci. Technol., A4 (1986), p. 2695.

13. O. Knotek et al.. Mat. Sci. Eng., A105/106 (1988), p. 481.

14. P. Duwez and F. Odell, J. Electrochem. Soc., 97 (1950), p. 299.

15. Y.L. Kuo et al., Electrochemical and Solid-State Letters, 7 (3) (2004), pp. C35-C37.

16. C.K. Hu and J.M.E. Harper, Mater. Chem. Phys., 52 (1998), p. 5.

17. "Powder Diffraction File" (Philadelphia, PA: Joint Committee on Powder Diffraction Standards of ASTM, 1996), Card 38-1420.

18. "Powder Diffraction File" (Philadelphia, PA: Joint Committee on Powder Diffraction Standards of ASTM, 1996), Card 32-1284.

19. L.E. Thod, Transition Metal Carbides and Nitrides (New York: Academic, 1971).

20. Y.L. Kuo et al., J. Electrochem. Soc., 151 (3) (2003), p. C176.

21. P.E. Schmid, M.S. Sunaga, and F. Levy, J. Phys. Chem. Solids, 30 (1969), p, 1835.

22. W.D. Callister, Jr., Fundamentals of Materials Science and Engineering (New York: Wiley, 2001).

23. Y.L. Kuo et al., Mater. Chem. Phys., 80 (3) (2003), p. 690.

24. "Powder Diffraction File" (Philadelphia, PA: Joint Committee on Powder Diffraction Standards of ASTM, 1996), Card 23-224.

25. J.S. Chen and J.L. Wang, J. Electrochem. Soc., 147 (5) (2000), p. 1940.

Chiapyng Lee and Yu-Lin Kuo are with the Department of Chemical Engineering at the National Taiwan University of Science and Technology inTaipei 106, Taiwan, ROC. Chiapyng Lee can be reached at +8862-2737-6623; fax +886-2-2737-6644; or e-mail [email protected] ch.ntust.edu.tw.

Copyright Minerals, Metals & Materials Society Jan 2007

(c) 2007 JOM. Provided by ProQuest Information and Learning. All rights Reserved.