Journal of The Electrochemical Society, 151 6 C369-C374 2004 0013-4651/2004/151 6 /C369/6/$7.00 The Electrochemical Society, Inc. Influence of Oxide Thickness on Nucleation and Growth of Copper on Tantalum Aleksandar Radisic,* Gerko Oskam,*,a and Peter C. Searson**,z Department of Materials Science and Engineering, Johns Hopkins University, Baltimore, Maryland 21218, USA C369 Oxide layers can play an important role in determining the kinetics of nucleation and growth during electrodeposition. In this paper, we report on the influence of oxide thickness on the nucleation and growth of copper on tantalum. Anodic oxidation was used to grow Ta 2 O 5 films up to about 20 nm thick. Current-voltage curves for anodically oxidized tantalum films show features associated with the nucleation and growth of copper but no stripping peaks as expected for a rectifying oxide film. Current transients for deposition of copper follow the rate law for instantaneous nucleation followed by diffusion-limited growth, independent of oxide thickness; however, the island density decreases exponentially with oxide thickness. 2004 The Electrochemical Society. DOI: 10.1149/1.1738673 All rights reserved. Manuscript submitted September 9, 2003; revised manuscript received January 6, 2004. Available electronically April 30, 2004. Current copper metallization technology 1-3 involves deposition of a diffusion barrier layer and a copper seed layer prior to electrochemical deposition of the copper. The diffusion barrier, typically Ta, TaN, TiN, or TiSiN, isolates the metallization from the silicon dioxide layer and prevents diffusion into the dielectric. As the feature sizes of trenches and vias continue to shrink, deposition of a continuous defect-free copper seed layer becomes increasingly difficult. As a result, strategies for direct deposition on the barrier layer are being explored. On exposure to air, tantalum forms a thin oxide layer that is very stable and can only be dissolved in concentrated HF. Thus, the oxide layer is expected to play a key role in determining the energetics and kinetics of copper deposition onto diffusion barrier layers. For systems where the deposition occurs by island growth, the critical thickness for island coalescence is inversely proportional to the square root of the island density. Consequently, to ensure void-free deposition in patterned structures, the critical thickness for island coalescence must be significantly smaller than the feature size. In this paper, we report on the nucleation and growth of copper on Ta as a function of oxide thickness. Oxide films up to 20 nm thick were grown by anodic oxidation. We show that for deposition of copper from sulfate solution, the mechanism of deposition is independent of oxide thickness, although the island density decreases exponentially with oxide thickness. Experimental Ta 2 O 5 films were grown electrochemically on Ta foil Alfa Aesar. Prior to all experiments, the Ta foil was rinsed in acetone, isopropanol, and distilled, deionized DI water. Ta 2 O 5 films were grown in 0.1 M H 2 SO 4 at constant potential using a two-electrode configuration, with a rectangular (14 8 cm) titanium counter electrode and a tantalum working electrode positioned 5 cm apart. At each potential, growth was terminated when the potential drop over a 100 resistor connected in series with the working electrode reached a steady-state value. After the oxide film was grown, the samples were rinsed with distilled, DI water and dried in air. Copper was deposited on the Ta 2 O 5 films at both constant potential and constant current from a solution containing 50 mm CuSO 4 5H 2 O and 375 mm H 2 SO 4. All the deposition experiments were performed using a three-electrode arrangement with a platinum mesh counter electrode and a Ag/AgCl reference electrode 0.22 V vs. normal hydrogen electrode. All potentials are reported with respect to Ag/AgCl electrode. The working electrode area was 1.5 cm 2. Impedance measurements were performed at 0.5 V vs. Ag/ AgCl with a 10 mv root-mean-square sinusoidal perturbation in the frequency range from 10 khz to 1 Hz. Results and Discussion Characterization of oxide film thickness. Ta 2 O 5 films were grown by anodic oxidation at potentials from 1 to 40 V in 0.1 M H 2 SO 4. The anodic oxidation of Ta results in the formation of tantalum pentoxide through the following reaction 4,5 2Ta 5H 2 O Ta 2 O 5 10H 10e The current efficiency for this reaction is essentially 1 because the rates of parasitic reactions such as oxygen evolution are negligible. Consequently, the change in oxide thickness d( d d 0 ) can be determined from d qv m zf where q is the charge passed during anodic oxidation, V m is the molar volume, z is the number of electrons in the overall reaction, and F is Faraday s constant. Because it is not possible to maintain an oxide-free surface in aqueous solution, the change in oxide thickness is referenced to the initial thickness of the air-formed film d 0. 6 Figure 1 shows the change in film thickness vs. anodic oxidation potential. The change in film thickness was determined from the charge passed during oxidation, taking V m 50.6 cm 3 mol 1 and z 10. 1 2 * Electrochemical Society Student Member. ** Electrochemical Society Active Member. a Present address: Departamento de Fisica Aplicada, Centro de Investigacíon de Estudios Avanzados-Instituto Politécnico Nacional, Unidad Merida, Merida, Yucatan 97310, Mexico. z E-mail: searson@jhu.edu Figure 1. Change in oxide thickness ( d) vs. anodic oxidation potential.
C370 Journal of The Electrochemical Society, 151 6 C369-C374 2004 Figure 4. Reciprocal capacitance, determined from the impedance measurements, vs. the change in oxide thickness ( d), determined from the charge passed during anodic oxidation. Figure 2. Bode plots for air-formed oxide and for oxide films grown at 10 V, 20 V, and 40 V. In all cases, the impedance response was measured at 0.5 V. Figure 2 shows Bode plots and phase angle recorded at 0.5 V in 0.1 M H 2 SO 4 for an air-formed film along with films grown at 10, 20, and 40 V. The Bode plots show capacitive behavior over the measured frequency range with slopes between 0.92 and 0.95. The thickness of the anodic oxide layer can be determined from capacitance measurements. The oxide layer can be considered a parallel plate capacitor with C ox 0 /d where is the relative permittivity, 0 is the permittivity of free space, and d is the oxide thickness. The oxide capacitance C ox and double-layer capacitance C H are in series; thus 1/C 1/C ox 1/C H, where C is the measured capacitance. 7 Thus, as long as C ox C H, then 1/C 1/C ox. At sufficiently positive potentials, up to the oxide formation potential, C ox is independent of potential, however, at more negative potentials C ox increases. The dependence of C ox on potential in this regime can be approximated by the Mott-Schottky equation for an n-type semiconductor. 6 Consequently, at negative potentials C ox may become of similar magnitude to C H resulting in an increase in potential drop across the Helmholtz layer. Figure 3 shows the reciprocal capacitance vs. anodic oxidation potential. The capacitance was determined from Bode plots recorded at 0.5 V in 0.1 M H 2 SO 4 over the frequency range 10-100 Hz. Figure 4 shows the reciprocal capacitance plotted vs. the change in oxide thickness determined from the charge passed during anodic oxidation. The linear dependence confirms that 1/C 1/C ox. From the slope, we determine a value for the relative permittivity of 40.3, slightly larger than the values of 25-30 reported in the literature. 8 Assuming that the increase is related to surface roughness, we obtain a roughness factor of about 1.3. Extrapolation of the linear region to the abscissa gives a value for the initial oxide thickness, prior to anodic oxidation, of 1.0 nm. Nucleation and growth of copper. Figure 5 shows cyclic voltammograms CVs for Ta in 50 mm CuSO 4 5H 2 O and 375 mm H 2 SO 4 with different oxide thicknesses. For the air-formed oxide film (d 1.0 nm), the onset of copper deposition occurs at about 0.25 V and is followed by a characteristic peak associated with nucleation and diffusion-limited growth at about 0.42 V. 9 The Figure 3. Reciprocal capacitance vs. anodic oxidation potential. Figure 5. CVs for deposition of copper on Ta with a Ta 2 O 5 oxide layer thickness of 1.0 nm air-formed film and 7.6, 12.8, and 20 nm at a scan rate of 10 mv s 1.
Journal of The Electrochemical Society, 151 6 C369-C374 2004 C371 Figure 6. Copper deposition current density at 0.42 V, determined from CVs, vs. anodic oxide thickness. deposition peak is followed by the onset of hydrogen evolution at 0.75 V. In the reverse scan, the diffusion-limited deposition current is 3.5 ma cm 2. Copper deposition continues up to 0.05 V, 0.47 V positive to the onset potential, indicating a very large nucleation overpotential. At more positive potentials, a copper stripping peak of 4mAcm 2 is seen at 0.12 V. The charge associated with anodic oxidation of Ta is negligible in this potential range; hence, the stripping charge of 87 mc cm 2 corresponds to approximately 145 equivalent monolayers of copper. For samples with anodically grown oxides, the copper deposition peak is shifted to more negative potentials and decreases with increasing oxide thickness. On the reverse scan, the diffusion-limited copper deposition current decreases slightly with increasing oxide thickness, and at more positive potentials, no stripping peak is observed. These features are similar to voltammograms obtained for copper deposition on TaN. 10 Diode-like behavior, where only reduction currents are observed, is typical of reactions on oxidized metals such as tantalum, zirconium, and aluminum. 7 Figure 6 shows the copper deposition current density at 0.42 V, from CVs, vs. oxide thickness. This potential corresponds to the kinetic regime for copper deposition on anodically grown oxide films. The current density decreases exponentially with increasing oxide thickness, consistent with a tunneling process. The apparent tunneling constant of 0.039 nm 1 is relatively small indicating that the mechanism involves hopping between states in the oxide and is not due to direct tunneling from electrons in tantalum to Cu II states in solution. 7 Figure 8. Reduced parameter current transients for deposition of copper on Ta at 0.75 V with a Ta 2 O 5 oxide layer thickness of a 1.0 nm air-formed film, b 1.3 nm, and c 7.6 nm. Figure 7. Current transients for deposition of copper on Ta at 0.75 V with ata 2 O 5 oxide layer thickness of 1.0 nm air-formed film and 1.3, 7.6, 12.8, and 20 nm. The kinetics of copper nucleation and growth on oxidized Ta films were determined by analysis of deposition transients. Figure 7 shows a series of current transients for copper deposition for different oxide thicknesses in 50 mm CuSO 4 5H 2 O and 375 mm H 2 SO 4 solution. In all cases, the potential was stepped from the open-circuit potential, where no deposition is observed, to 0.75 V. For oxide thicknesses of 1,0, 1.3, 7.6, and 12.8 nm, about 120-330 mc cm 2 200-550 equivalent monolayers of copper were deposited; for the 20 nm thick oxide 20 mc cm 2 34 equivalent monolayers were deposited. The current transients are characterized by an initial increase in current due to nucleation and threedimensional 3D diffusion-limited growth of isolated clusters. At longer times, the current decreases due to the transition from 3D to one-dimensional diffusion-limited growth. The current transients shown in Fig. 7 are replotted in Fig. 8 in
C372 Journal of The Electrochemical Society, 151 6 C369-C374 2004 Figure 10. Logarithm of island density vs. oxide thickness for deposition of copper on Ta at 0.75 V. Air-formed film and anodically oxidized films. Figure 9. SEM images of copper islands deposited on Ta at 0.75 V with a Ta 2 O 5 oxide layer thickness of a 1.0 nm air-formed film, b 1.3 nm, c 7.6 nm, and d 12.8 nm. In all cases, 200-550 equivalent monolayers of copper were deposited. dimensionless form and compared to the growth laws for instantaneous and progressive nucleation followed by diffusion-limited 3D growth. 11,12 For instantaneous nucleation, the time-dependent deposition current density normalized to the geometric surface area is given by 11,12 i t zfc 0D 1/2 1/2 t 1/2 1 exp N 0 D 8 3 c 0 V m 1/2 t 3 the model for instantaneous nucleation followed by 3D diffusionlimited growth. For oxide thicknesses greater than 1.3 nm, additional peaks were seen at longer times; such peaks are often associated with renucleation. 13 These results indicate that the mechanism of nucleation and growth of copper on tantalum oxide is independent of oxide thickness. Figure 9 shows plan view scanning electron microscope SEM images of copper clusters deposited at 0.75 V on Ta films with different oxide thicknesses. These images show that the cluster density decreases with increasing oxide thickness. It is also evident that clusters deposited on thick oxide layers exhibit complex shapes due to renucleation. For d ox 12.8 nm, no copper clusters were detected on the substrates. The island density, obtained from SEM images, decreases exponentially with increasing oxide thickness for anodically grown oxides, as shown in Fig. 10. The island density for the air-formed film is about an order of magnitude larger than expected from extrapolation of the exponential fit to the results obtained from the anodically grown oxides. This effect is due to defects in the air-formed film which give rise to a spatially inhomogeneous current distribution. 14 The inverse slope (dd ox /d log N) of the island density for the anodically grown oxide films is 9.0 nm/decade. The island densities obtained from analysis of the transients using the Sharifker-Hills model also show an exponential dependence on the oxide thickness, although the values are about two orders of magnitude lower. The increase in oxide thickness results in an increase in the potential drop across the oxide ( U ox ) and hence a decrease in the where c 0 is the bulk concentration, D is the diffusion coefficient, z is the valence of the metal ion, F is Faraday s constant, N 0 is the nucleus density, and V m is the molar volume of the solid. The normalized current density for instantaneous nucleation followed by diffusion-limited growth is given by i 2 2 i max 1.9542 t max t 1 exp 1.2564 t 2 t max 4 From Fig. 8 it can be seen that the deposition transients show good agreement with the rate law for instantaneous nucleation followed by diffusion-limited growth. The experimental data for d ox 1.0 nm air-formed film, 1.3 nm, and 7.6 nm compare well with Figure 11. Potential transients for deposition of copper on Ta at 3 ma cm 2 withata 2 O 5 oxide layer thickness of 1, 1.3, 7.6, 12.8, and 20 nm.
Journal of The Electrochemical Society, 151 6 C369-C374 2004 C373 Figure 13. Logarithm of copper island density vs. oxide thickness for copper deposited at 3 macm 2 for 30 s 150 equivalent monolayers. Air-formed film and anodically oxidized films. Figure 12. SEM images of copper islands deposited on Ta at 3 macm 2 for30s 150 equivalent monolayers withata 2 O 5 oxide layer thickness of a 1.0 nm air-formed film, b 1.3 nm, c 7.6 nm, and d 12.8 nm. potential drop across the Helmholtz layer ( U H ). The decrease in the potential drop across the Helmholtz layer reduces the driving force for nucleation and growth. For a dielectric film where 1/C ox d, the potential drop in the oxide ( U ox ) is expected to be proportional to the oxide thickness. Because the total potential drop across the interface U total U ox U H, the driving force for nucleation and growth ( U H ) is inversely proportional to the oxide thickness. The exponential dependence of cluster density on oxide thickness at constant potential is equivalent to an exponential dependence on applied potential at constant oxide thickness. In previous work, we have shown that the island density for deposition of copper on TaN as well as on other substrates increases exponentially with decreasing potential with an inverse slope (du/d log N) of about 100 mv/decade. 10 Thus, an order of magnitude increase in island density can be achieved by decreasing the oxide thickness by 9.0 nm or by decreasing the applied potential by 100 mv. Electrodeposition of copper metallization is usually carried out in a two-electrode arrangement at constant current. However, nucleation and growth processes are more difficult to study under conditions of constant current because the driving force supersaturation is not constant. Potential transients in response to a current step are characterized by an increase at short times due to charging and nucleation of copper clusters. 15,16 As the number of stable clusters increases with time, the potential reaches a maximum and then decreases to values at which no new stable clusters are formed. 16,17 Figure 11 shows potential-time transients for copper deposition in response to a current step of i 3mAcm 2 for 30 s. Generally, the overpotential is larger for deposition on thicker oxide layers due to the larger potential drop across the oxide layer. Note that the potential for deposition on oxide layers 12.8 and 20 nm thick is in the region where hydrogen evolution is seen in the current-potential curves see Fig. 5. Figure 12 shows plan view SEM images of copper islands deposited at constant current. The island density is highest for d ox 1.0 nm air-formed film and decreases with increasing oxide thickness. For d ox 12.8 nm, the islands are approximately hemispherical but are composed of multiple particles. No copper clusters were observed on samples d ox 20 nm. Figure 13 shows the copper island density vs. oxide layer thickness for deposition at 3 macm 2 for 30 s. As for the case of deposition at constant potential Fig. 10, the island density decreases exponentially with increasing oxide thickness for the anodically grown oxides. The island density for the air-formed film is about four times larger than would be expected from extrapolation of the results for the anodically grown oxides. The inverse slope of the island density for the anodically grown oxides is 18.2 nm/ decade. A factor of two larger than that obtained for deposition at constant potential. Conclusion Copper deposition on oxidized tantalum follows the rate law for instantaneous nucleation followed by 3D diffusion-limited growth, independent of oxide thickness. Increasing the oxide thickness results in an exponential decrease in island density. This effect is related to the decrease in potential drop across the Helmholtz layer. Acknowledgment The authors acknowledge support from the National Science Foundation grant no. CTS-0122909.
C374 Journal of The Electrochemical Society, 151 6 C369-C374 2004 Johns Hopkins University assisted in meeting the publication costs of this article. References 1. D. Edelstein, J. Heidenreich, R. Goldblatt, W. Cote, C. Uzoh, N. Lustig, P. Roper, T. McDevitt, W. Motsiff, A. Simon, J. Dukovic, R. Wachnik, H. Rathore, R. Shultz, L. Su, S. Luce, and J. Slattery, Tech. Dig. - Int. Electron Devices Meet., 1997, 773. 2. D. C. Edelstein, G. A. Sai-Halasz, and Y. J. Mii, IBM J. Res. Dev., 39, 383 1995. 3. P. C. Andricacos, C. Uzoh, J. O. Dukovic, J. Horkans, and H. Deligianni, IBM J. Res. Dev., 42, 567 1998. 4. L. Young, Anodic Oxide Films, Academic Press, New York 1961. 5. H. Bispinck, O. Ganschow, L. Wiedmann, and A. Benninghoven, Appl. Phys., 18, 113 1979. 6. V. Macagno and J. W. Schultze, J. Electroanal. Chem. Interfacial Electrochem., 180, 157 1984. 7. S. R. Morrison, Electrochemistry at Semiconductor and Oxidized Metal Electrodes, Plenum, New York 1980. 8. D. J. Smith and L. Young, Thin Solid Films, 101, 11 1983. 9. E. Budevski, G. Staikov, and W. J. Lorentz, Electrochemical Phase Formation and Growth, VCH, New York 1996. 10. A. Radisic, Y. Cao, P. Taephaisitphongse, A. C. West, and P. C. Searson, J. Electrochem. Soc., 150, C362 2003. 11. G. Gunawardena, G. J. Hills, I. Montenegro, and B. R. Scharifker, J. Electroanal. Chem. Interfacial Electrochem., 138, 225 1982. 12. B. R. Scharifker and G. J. Hills, Electrochim. Acta, 28, 879 1983. 13. Southampton Electrochemistry Group, Instrumental Methods in Electrochemistry, Ellis Horwood Limited, New York 1985. 14. S. B. Basame and H. S. White, Anal. Chem., 71, 3166 1999. 15. D. Kashchiev, Thin Solid Films, 29, 193 1975. 16. M. Peykova, E. Michailova, D. Stoychev, and A. Milchev, Electrochim. Acta, 40, 2595 1995. 17. A. Milchev and M. I. Montenegro, J. Electroanal. Chem., 333, 93 1992.