Numerical and Experimental Analysis of Cu Diffusion in Plasma-treated Tungsten Barrier

To investigate the effects of N2O plasma treatment on W barrier, chemical bonding states of barrier were analyzed by XPS. Figures 5.11(a), (b), and (c) show the W 4f, O 1s, and N 1s spectra of as-deposited and N2O plasma-treated W barriers.

The W 4f7/2 and W 4f5/2 peaks are situated at the same position (31.2, 33.4 eV) for two barriers, and the two peaks are characteristics of the W itself [23]. However, two another peaks at binding energies 36.16 and 38.43 eV are observed for the W(N2O) barrier. The peaks correspond to binding energies of tungsten oxide and are attributed to oxidation during N2O plasma treatment. The O 1s spectrum of the W(N2O) barrier exhibits a strong and broad peak centered at around 531.2 eV, as shown in Fig. 5.11(b). There is almost no peak in untreated W barrier. W-O phases are believed to form on the surface of W(N2O) barrier. The N 1s peak of the W(N2O) barrier can be well resolved into two peaks by curve fitting compared to no peak in the N 1s spectrum of the W barrier, as shown in Fig. 5.11(c). These two peaks are centered at 397.5 and 399.25 eV. The weak peak centered at 397.5 eV is consistent with the N 1s binding energy of nitride compound and is also reported by previous research [25]. Another strong peak at ~399.25 eV is observed in N 1s spectrum and attributed to the N atoms or molecules present in grain boundaries of W [13,14,30]. It indicates that some N atoms do not form strong covalent or ionic bonds with W during N2O plasma treatment.

Figure 5.12(a) displays cross-sectional bright field TEM micrograph of as-sputtered W film. Columnar grain structure is observed. Cu diffusion in columnar grains is believed to be relatively easy because columnar grains will provide fast diffusion paths for Cu diffusion. Figures 5.12(b) and (c) display plane-view TEM

images and selected area diffraction (SAD) patterns of W and W(N2O) barriers.

Several sharp ring is observed for untreated W film. It indicates that the as-deposited W film is a polycrystalline structure. The grain size of as-deposited W film is 20-40 nm, as shown in Fig. 5.12(b). The W(N2O) barrier has a diffused ring pattern instead of diffraction spots, indicating that the N2O plasma treatment causes an amorphous surface layer upon W film. Grain size of the W(N2O) barrier is only 2-5 nm. It indicates that a surface layer with finer grains is formed due to the reactions and bombardments of energetic radicals and ions during plasma treatment. It is reported that the nanostructured amorphous diffusion barrier, defined as a very short-range order single crystal, is highly attractive due to its relatively thermal stability and its relatively higher resistance against Cu diffusion [33].

Figure 5.13 displays resistivity and root-mean-square (RMS) surface roughness of plasma treated W film as a function of plasma treatment time. Surface roughness reduces apparently after plasma treatment, as shown in Fig. 5.13. Plasma treatment could sputter the barriers and make them smooth. Sputtering and stuffing effects are believed to occur due to the reactions or bombardments of energetic radicals and ions during plasma treatment. Resistivity increases with increasing plasma treatment time.

It is expected that resistivity of W with plasma treatment would increase because a high-resistivity surface layer is formed and thus the effective thickness of the W film with low resistivity reduces. Resistivity of W(N2O) barrier is higher than those of the W(N2) and W(NH3) barriers. High resistivity may be due to the formation of high-resistance W-O compounds in W(N2O) barrier, as indicated in XPS analyses.

However, the resistivity of W(N2O) (~24 μΩ cm) barrier is still much lower than that of the reactively sputtered WN film (150~200 μΩ cm) [14].

Table 5.2 summarizes the properties of W, WN and W(N2O) barriers. Table 5.2 also lists properties of sputtered Ta and Ti films in literatures [13,31,32,34] for

comparison. The resistivities of sputtered W and W(N2O) barriers are about 20 and 24 μΩ cm, respectively. The resistivities are very close and much lower than that of the reactively sputtered WN (150-200 μΩ cm) [14], Ta (~160 μΩ cm) [13,31], and Ti (~70 μΩ cm) [34]. Moreover, it is reported that electroplating of Cu layer is difficult on seed layers with rough surface [35]. Surface roughness could be reduced from 2.5 to 0.5 nm by N2O plasma treatment, as listed in Table 5.2. Grain size is calculated from plane-view TEM. The as-deposited W film is columnar grain structure with a grain size of 20-40 nm and grain size of the W(N2O) barrier is only 2-5 nm. It indicates that nanocrystallization effect would occur due to the reactions and bombardments of energetic radicals and ions during N2O plasma treatment.

Furthermore, it is reported that the impurities (nitrogen or oxygen) in the films are responsible for the intrinsic compressive stress [21]. In this research, the tensile stress decreases from 1.9×10¹⁰ to 1.4×10⁹ dynes/cm² as N2O plasma treatment is applied to W film, as listed in Table 5.2. N2O plasma treatment also enhances the adhesion between Cu and W(N2O) barrier. The adhesion strength of Cu on W(N2O) barrier (50-58 Mdynes/cm²) is better than untreated W (30-35 Mdynes/cm²) and other barriers such as Ta and Ti [32]. It is suggested that O or N species on the barrier surface can react with Cu to form more stable interface and thus promote adhesion after N2O plasma treatment.

Figure 5.14(a) shows the variation percentage of the Cu/barrier/Si system after furnace annealing at various temperatures. The variation in sheet resistance is defined as the ratio of (R-R0) to R0, in which R0 and R denote the sheet resistance of as-deposited and annealed samples. The results reflect the interactions between Cu and Si indirectly. Resistance increases rapidly at certain temperature because of failure of the diffusion layer and formation of compounds. Sheet resistance of Cu/W(N2O)/Si increases slightly even after annealing at 750°C. However, sheet

resistances of samples with other barriers (W, WN [14], WCx [27], TiCx [28], and TaN [29]) sharply increase after annealing at 600-700°C, indicating that a considerable amount of Cu has already diffused through the barrier layers and resulted in Cu3Si compounds, and thus strongly deteriorated the conductivity of the contact system.

X-ray diffraction technique is further used to detect the structural change in the annealed samples. Figure 5.14(b) shows XRD patterns of Cu/W/Si and Cu/W(N2O)/Si systems after annealing at 700°C for 30 min. Peaks of Cu3Si compounds are observed for W barrier and the resulting XRD pattern is consistent with variation in sheet resistance. There is no Cu3Si peak in W(N2O) barrier system, indicating that W(N2O) has an excellent barrier performance. It is noted that there is no Cu-W compound in the XRD spectra, and similar results are found for WN barriers [14]. The failure of W and WN barriers is attributed that Cu atoms diffuse through defects and grain boundaries of the barrier without reacting with the WN and W film.

Failure temperature (variation percentage > 50%) of W(N2O) barrier is higher than 750°C and superiors to sputtered WCx(~650°C) [27], TiCx(~625°C) [28], and TaN(~650°C) [29] barriers.

Barrier performance is also evaluated by the leakage current density of the junction diode. Figure 5.15 illustrates the statistical distributions of leakage current densities of Cu/barrier/n⁺-p junction diodes measured at reverse bias of 5 V after annealing at 500 and 600°C. The leakage current densities of all diodes are below 10^-8 A/cm² before annealing. Most diodes with sputtered W, WN [14], WCx [27] and TiCx

[28] barriers have leakage current densities higher than 1×10^-7 A/cm² after annealing at 500°C for 30 min. Diodes with W(N2O) barriers retain leakage current densities less than 10^-8 A/cm² after annealing at 500°C. Moreover, most diodes with W(N2O) barriers retain leakage current densities less than 10^-8 A/cm² even after annealing at 600°C. W(N2O) show an excellent barrier capability against Cu diffusion compared to

W, WN [14], WCx [27], TiCx [28] and TaN [29], as shown in Fig 5.15.

Although barrier performance of the W film is significantly improved by N2O plasma treatment, similar plasma treatments, e.g., N2 plasma treatment, had been proposed to enhance barrier performance in previous researches [13,30]. W barriers are also treated by N2 or NH3 plasma for comparison in the work. Figure 5.16 exhibits statistical distributions of leakage current densities of Cu/W(N2O)/n⁺-p, Cu/W(N2)/n⁺-p and Cu/W(NH3)/n⁺-p diodes after annealing at 500 and 600°C. All W(N2O), W(N2), and W(NH3) show an enhanced barrier performance compared to untreated W. Moreover, N2O plasma treatment has a better barrier improvement than N2 plasma treatment and N2 plasma is more effective than NH3 plasma. It is found that some diodes have leakage current densities higher than 1×10^-7 A/cm² for W(NH3) barriers after annealing at 500°C. Most diodes with W(N2O) barriers retain leakage current densities less than 1×10^-8 A/cm² and large leakage current densities in range of 10^-7 to 10^-4 are found for Cu/W(N2)/n⁺-p and Cu/W(NH3)/n⁺-p diodes after annealing at 600°C. Enhancing performance of the W(N2O) barrier is attributed to combined effects of nitridation and oxidation during N2O plasma treatment. N2O is a strong oxidizing agent and energetic oxygen radicals and atoms are more easily produced by dissociation of N2O. In contrast, it is less effective to dissociate triple bonds of nitrogen as nitrogen plasma is used to post-treat W barrier. The bond strengths of N-O and N-N are 630.57 and 945.33 kJ/mol. Energetic oxygen radicals and atoms are helpful in formation of an oxygen stuffed layer on the columnar-grained W barrier. It will act as a more efficient barrier against the Cu diffusion. N2 plasma treatment doesn’t induce oxygen stuffed layer on surface of the W barrier and nitrogen radicals and atoms are limited for nitridation and stuffing compared to N2O plasma treatment.

Bond strength of N-H is 339 kJ/mol for NH3 plasma treatment. Although nitridation is expected to occur greatly, no plasma oxidation on W surface occurs and lots of atomic

hydrogen will diffuse into the grain boundaries of W due to effective dissociation of NH3. Hydrogen will out-diffuse from grain boundaries of plasma-treated W after annealing at a certain temperature, lead to defect sites or carrier trapping sites, and hence, decrease its resistance to copper penetration.

Barrier capabilities are further investigated by evaluating Cu diffusion in W and W(N2O) barriers. The diffusion of Cu in the temperature range of 600-700°C is evaluated by SIMS and the penetration depth profiles of specimens are indicated in Fig. 5.17. Relatively low Cu penetration into the W(N2O) barrier is detected compared to sputtered W barrier. Since the mixing enthalpy of Cu solute in W is large (8.0×10⁴ j/mol), Cu atoms do not intermix with W at high temperature [36,37]. As far as we know, it is relatively difficult for Cu penetration into W barrier by lattice diffusion. However, the diffusion coefficients of Cu in grain boundaries of W barrier are still unknown, though the rough estimation has been discussed [38], and the results are still needed to be further discussed.

Whipple had given formulae for the concentration in a semi-infinite region of low diffusion coefficient bisected by a thin well-diffusing slab. Since our source condition is close to infinite one, the Whipple’s solution is used for profile analysis [17]. The key result of Whipple’s solution is

2 surface concentration, t is the annealing time, and δ is the grain boundary width. DL

and DB present the diffusion coefficients in the lattice and the grain boundary. Figure 5.18 displays the concentration profiles obtained from diffusion of Cu in W barriers at

various temperatures in the standard coordinates lnC vs. ⁵

6

y . Based on Eq. 5.1 and provided that DL is known, δDB can be determined by measuring the slope from the linear region in the lnC vs. ⁵

6

y plots. The value of DL can be found from the initial part of the concentration profile. Under the assumptions of a semi-infinite system and constant source at the surface, the boundary conditions applied to solve Fick’s diffusion equation are

C= Cs, y < 0, t = 0 (5.2a)

C = 0, y > 0, t = 0 (5.2b)

An error function solution can be expressed by the equation

⎟⎟

where the error function was fitted to the initial part of the concentration profiles and extrapolated to zero thickness y = 0 in the case of subtracting the grain boundary contribution. The results are shown in Table 5.3, and the derived value of DL was used to calculate δDB. To obtain a value DB, a grain boundary width must be assumed, it is reasonable to assume a width of about two atom layers, δ = 5×10^-8 cm [39,40].

Based on the numerical calculation, a satisfactory fitting to the experimental depth profile can be obtained [16]. The solid curves shown in Fig. 5.18 are the fitted curves according to the derived parameters. The temperature dependence of the grain boundary diffusivity in W barrier can be expressed by the Arrhenius relation of DB = DB0 exp(-QB/kT), as plotted in Fig. 5.19, where DB0 is the pre-exponential factor, QB is the activation energy for grain boundary diffusion, k is the Boltzmann constant, and T is annealing temperature. Also, the temperature dependence of the lattice diffusivity in W barrier can be expressed by the Arrhenius relation of DL = DL0 exp(-QL/kT), where DL0 is the pre-exponential factor and QL is the activation energy for lattice diffusion.

The dependency is plotted in Fig. 5.20. The values of diffusivities, pre-exponential factors, and activation energies are summarized in Table 5.3. The D values of Cu in sputtered W films are smaller than those in CVD-W films, but the magnitudes are in the same order. In addition, the activation energy of Cu diffusion in sputtered W films is somewhat larger than that in CVD-W films [38,41].

Similar numerical analysis can be applied to evaluate the effects of Cu diffusion in W(N2O) barriers. As mentioned previously, N2O plasma treatment causes an amorphous surface layer upon the W film. This is the case of a semi-infinite medium which has a skin or surface layer W(N2O)1 having diffusion properties different from those of the rest of the medium W(N2O)2. The subscripts 1 and 2 denote the amorphous surface layer and the rest in the W(N2O) barrier. Thus, suppose in the semi-infinite region –h < y’ < ∞, the diffusion coefficient is DL1 in the region –h < y’

< 0, and the concentration is denoted by C1 there, while the corresponding quantities in y’ > 0 are DL2 and C2. Assume the conditions at the interface to be

the solution to the problem of zero initial concentration and the surface y’ = -h maintained at constant concentration C0 is given as following [18]

( ) ( )

The thickness h of the surface layer W(N2O)1 is about 3 nm from high-resolution TEM micrograph. Both diffusion coefficients can be obtained by numerical calculation. The values of DL1 and DL2 can be roughly estimated from fitting of experimental concentration profiles using an error function solution such as that in Eq.

5.3. The estimated values are used as the initial guess of DL1 and DL2 and further substituted into the Eqs. 5.4 and 5.5 to obtain concentrations C1 and C2, respectively.

To quickly obtain the numerical convergence, the high order terms are neglected in the Eqs. 5.4 and 5.5 during the calculation. It is found that the calculated error is smaller than 10^-4 as n = 4. The relatively exact lattice diffusion coefficients can be found by fitting of the experimental concentration profile. Figure 5.21 shows experimental and calculated concentration profiles. The derived values of DL1 and DL2

are listed in Table 5.4.

Similarly, grain boundary diffusion coefficients DB2 in W(N2O)2 region can be determined from the derived lattice diffusion coefficient DL2 and the slope in lnC vs.

5 6

'

y plots. Figure 5.22 displays experimental and calculated concentration profiles for Cu diffusion in W(N2O) barriers at various annealing temperatures in the standard coordinates lnC vs. ⁵

6

'

y . The derived grain boundary diffusion coefficients DB2 are listed in Table 5.4. Table 5.4 summaries the values of diffusion coefficients and pre-exponential factors for Cu diffusion in W(N2O) barriers. Other barrier materials in literatures also list for comparison [38,41,42]. The W(N2O) barrier shows small lattice diffusion coefficients compared to W2N, W, and TiB2 barriers. The variation of DL2

with temperature is slight for the W(N2O) barrier, indicating that the W(N2O) barrier has better thermal stability.

One significant finding in the present study is that some atomic nitrogen and oxygen will react with W, segregate at grain boundaries of W film as impurities, and

act as a stuffing agent to block fast diffusion path during N2O plasma treatment.

Nitrogen addition will stuff the grain boundaries of W and nitrify tungsten to form tungsten nitride, as shown in XPS analyses of Fig. 5.11(c). The same effect has been reported after forming of TiN barrier layer at the surface of Ti layer [43]. Figure 5.23 shows cross sections of the interfacial structures of Cu/W/Si and Cu/W(N2O)/Si samples before and after annealing,. The as-deposited W barrier has a columnar grain structure as shown in Fig. 5.23(a). The failure of W barrier is attributed to the Cu diffusion through the columnar W to form Cu3Si after annealing at 700°C for 30 min.

Grain boundary diffusion coefficients are 2.3~4.7×10^-16 cm²/s using Whipple analysis of grain boundary diffusion. The barrier capability of the W film against Cu diffusion can be improved by N2O plasma treatment. An oxidized and nitrided layer with nanostructured grains is formed on the surface of the stuffed W barrier, as shown in Fig. 5.23(b). Relatively low diffusion coefficients are found. No Cu silicide compound is observed for Cu/W(N2O)/Si sample after annealing at 700°C for 30 min because nanostructured and stuffed barrier can effectively impede Cu diffusion.

5.4 Summary

Effects of N2O plasma treatment on the thermal stability of the Cu/WNx/n⁺-p junction system were systematically investigated. With N2O plasma treatments on WNx barriers, the Cu/WNx(N2O)/Si systems sustained thermal annealing up to 750°C without electrical degradation. N2O plasma treatment resulted in nitridation and oxidation of the WNx barrier and formation of a resulting amorphous layer on the surface. Furthermore, N2O plasma treatment introduces excessive nitrogen atoms to stuff grain boundaries of the WNx barrier, and hence, effectively suppresses the formation of Cu-Si compound and improves barrier performance.

The effectiveness of W(N2O) films as diffusion barriers between Cu and Si has been investigated. W(N2O) films, which have amorphous and nano-grained surface layers, show high thermal stability, low resistivity, low surface roughness, low tensile stress, and better adhesion with Cu. W(N2O) barriers show excellent barrier capabilities against Cu diffusion. Cu/W(N2O)/n⁺-p junction diodes retain leakage current densities less than 10^-8 A/cm² even after annealing at 600°C. Copper diffusion in W and W(N2O) barriers is further analyzed using the Whipple analysis of grain boundary diffusion and Fick’s diffusion law. Both lattice and grain boundary diffusivities of Cu diffusion in W and W(N2O) barriers are extracted from the Cu concentration profiles after annealing the samples at 600-700°C. Grain boundary diffusion coefficients of Cu in sputtered W films are 2.3~4.7×10^-16 cm²/s. Relatively low diffusion coefficients are found in W(N2O) barriers because oxidized and nitrided layers with nano grains are formed on the surface of the stuffed W barriers after N2O plasma treatments.

References

[1] C. A. Ross, Mater. Res. Soc. Symp. Proc. 225, 35 (1991).

[2] M. Iguchi, T. Takewaki, Y. Matsubara, Y. Kunimune, N. Ito, Y. Tsuchiya, T.

Matsui, K. Fujii, K. Motoyama, K. Sugai, A. Kubo, M. Suzuki, H. Tachibana, A.

Nishizawa, K. Nakabeppu, S. Yamasaki, S. Yokogawa, Y. Yamamoto, T. Kunugi, S. Nakata, M. Kagamihara, A. Shida, S. Nakamoto and H. Gomi, Technical Digest of the International Electron Devices Meeting (IEEE, Washington, 1999), p. 615.

[3] J. Torres, Appl. Surf. Sci. 91, 112 (1995).

[4] S. P. Murarka, Microelectron. Eng. 37/38, 29 (1997).

[5] H. Li, S. Jin, H. Bender, F. Lankmans, I. Heyvaent, K. Maex and L. Froyen, J.

Vac. Sci. Technol. B 18(1), 242 (2000).

[6] J. S. Becker and R. G. Gordon, Appl. Phys. Lett. 82, 2239 (2003)

[7] S. Wong, C. Ryu, H. Lee and K. Kwon, Mater. Res. Soc. Symp. Proc. 514, 75 (1998).

[8] S. Ganguli, L. Chen, T. Levine, B. Zheng and M. Chang, J. Vac. Sci. Technol. B 18(1), 237 (2000).

[9] M. Hecker, R. Hűbner, R. Ecke, S. Schulz, H. J. Engelmann, H. Stegmann, V.

Hoffmann, N. Mattern, T. Gessner and E. Zschech, Microelectr. Eng. 64, 269 (2002).

[10] P. J. Pokela, C. K. Kwok, E. Kowala, S. Raud, and M. A. Nicolet, Appl. Surf. Sci.

53, 364 (1991).

[11] Y. T. Kim, C. S. Kwon, D. J. Kim, J. W. Park and C. W. Lee, J. Vac. Sci. Technol.

A 16, 477 (1998).

[12] K. L. Ou, W. F. Wu, C. P. Chou, S. Y. Chiou and C. C. Wu, J. Vac. Sci. Technol. B 20, 2154 (2002).

[13] W. F. Wu, K. L. Ou, C. P. Chou and C. C. Wu, J. Electrochem. Soc. 150, G83 (2003).

[14] K. C. Tsai, W. F. Wu, J. C. Chen, T. J. Pan and C. G. Chao, J. Vac. Sci. Technol. B, 22, 993 (2004).

[15] W. F. Wu, K. C. Tsai, C. G. Chao, C. F. Huang, S. T. Wu, Y. L. Chin and B. S.

Chiou, in Proceedings of the 17^th International VLSI Multilevel Interconnection Conference, p. 490, Santa Clara, (June 2000).

[16] J. C. Lin and C. Lee, J. Electrochem. Soc., 146, 3466 (1999).

[17] R. T. P. Whipple, Philos. Mag., 45, 1225 (1954).

[18] J. Crank, The Mathematics of Diffusion, Clarendon Press, Oxford (1975).

[19] J. A. Thornton and D. W. Hoffman, Thin Solid Films 171, 5 (1989).

[20] A. G. Dirks, R. A. M. Wolters and A. E. M. De Veirman, Thin Solid Films 208, 181 (1992).

[21] H. Windischmann, J. Vac. Sci. Technol. A 9, 2459 (1991).

[22] B. S. Suh, H. K. Cho, Y. J. Lee, W. J. Lee and C. O. Park, J. Appl. Phys., 89(7), 4128 (2001).

[23] J. F. Moulder, W. F. Stickle, P. E. Sobol and K. D. Bomben, Handbook of X-ray Photoelectron Spectroscopy (Physical Electronics, Eden Prairie, MN, 1995).

[24] H. L. Zhang, D. Z. Wang and N. K. Huang, Appl. Surf. Sci. 150, 34 (1999).

[25] T. Nakajima, K. Watanabe and N. Watanabe, J. Electrochem Soc. 134, 3175 (1987).

[26] K. M. Chang, T. H. Yeh and I. C. Deng, J. Appl. Phys. 81 (8), 3670 (1997).

[27] S. J. Wang, H. Y. Tsai, S. C. Sun and M. H. Shiao, J. Electrochem. Soc. 148, 500 (2001).

[28] S. J. Wang, H. Y. Tsai, S. C. Sun and M. H. Shiao, J. Electrochem. Soc. 148, 563 (2001).

[29] M. H. Tsai, S. C. Sun, C. E. Tsai, S. H. Chuang and H. T. Chiu, J. Appl. Phys.

79(9), 6932 (1996).

[30] K. M. Chang, T. H. Yeh, I. C. Deng and C. W. Shih, J. Appl. Phys., 82, 1469 (1997)

[31] W. L. Yang, W. F. Wu, D. G. Lin, C. C. Wu and K. L. Ou, Solid-State Electron., 45, 149 (2001).

[32] H. Ono, T. Nakano and T. Ohta, Appl. Phys. Lett., 64, 1511 (1994).

[33] D. J. Kim, Y. T. Kim and J. W. Park, J. Appl. Phys., 82, 4847 (1997).

[34] M. E. Day, M. Delfino, J. A. Fair and W. Tsai, Thin Solid Films, 254, 285 (1995).

[35] T. Hara, Y. Yoshida and H. Toida, Electrochem. Solid-State Lett., 5, G36 (2002).

[36] M. E. Glicksman, Diffusion in Solid, p.207, Wiley Interscience, New York (2000).

[37] F. R. de Boer, R. Boom, W. C. M. Mattens, A. R. Miedema and A. K. Niessen, Cohesion in Metals, Transition Metal Alloys, North-Holland, Amsterdam (1988).

[38] M. Uekubo, T. Oku, K. Nii, M. Murakami, K. Takahiro, S. Yamaguchi, T.

Nakano and T. Ohta, Thin Solid Films, 286, 170 (1996).

[39] D. Gupta and T. T. C. Tsui, Appl. Phys. Lett., 17, 294 (1970).

[40] J. C. Fisher, J. Appl. Phys., 22, 74 (1951).

[41] K. Vieregge and D. Gupta, Tungsten and Tungsten Alloys-Recent Advances, p.231, The Minerals, Metals and Materials Society (1991).

[42] J. Pelleg and G. Sade, J. Appl. Phys., 91, 6099 (2002).

[43] T. Hara, K. Tani and K. Inoue, Appl. Phys. Lett., 57, 1660 (1990).

Table 5.1 Properties of tungsten nitride barriers deposited at various nitrogen flow ratios used in the study.

Table 5.1 Properties of tungsten nitride barriers deposited at various nitrogen flow ratios used in the study.